Skip to main content
Release history of the clawboo npm package, newest first. Each entry starts with what changed for you. The full notes for every release are in apps/cli/CHANGELOG.md, and Release process explains how a release is made.
These docs follow the main branch. The latest npm release is 0.9.0; anything on main that is newer is listed under Unreleased once it lands. Run clawboo --version to see your version, and npx clawboo@latest to run the latest release.

At a glance

Every 0.1.x release is deprecated on npm. That line ran on OpenClaw only and does not work reliably with current OpenClaw. Install a supported release with npm install -g clawboo@latest, or run npx clawboo@latest. Keep the @latest: a bare npx clawboo can reuse an old copy from npm’s cache.
2026-11-17
0.9.0 turns Home into a working page. Each team can keep goals, KPIs and pages, and Home shows how a team, or all of them, is doing, with every number saying where it came from. Behind a sign-in proxy, an administrator can now let each person’s role decide what they may do, let only requests the proxy signed act as a person, and give a program beside Clawboo a key of its own for its reads. The dashboard gets a new frame and a near-black dark mode, and Clawboo now stops the Hermes gateway the way Hermes expects.Home
  • Home is a working page. It is the first row in the sidebar’s nav, and its picker works like the Board’s: All teams, then each team. Clicking a team in the team rail opens Home on it. A team’s Home shows its status line, what needs you, its goals, its KPIs, the requests in flight, spend against its budget, upcoming routines, pinned pages and the files its agents made. All teams shows totals, one Needs you list for every team and a card per team. Every number says where it came from: computed from the board, reported by an agent (named, with its run), or typed by a person. See Home.
  • Goals. Each team can keep goals, each optionally under one parent goal, with an owner, a target date, a status and a health a person sets. A goal’s page shows the tasks linked to it, its KPIs, its updates and its cost. A task counts toward a goal when it or its request names one, and progress is done tasks over linked ones. Done and dropped goals fold under the goal list and can be reopened. See Goals.
  • KPIs. A person defines each KPI once: name, unit, target, period and where its value comes from. Its value can be computed from the board (such as tasks done, spend or run success rate), reported by an agent, or typed by a person. A value older than the KPI allows is marked stale. Agents can suggest a KPI, which a person accepts or dismisses. See KPIs.
  • Goal updates. Agents draft updates, and only a person publishes or discards one. A team’s Home settings can turn on a weekly team routine that asks the lead to draft an update for each active goal.
  • Pages. Markdown pages per team, linked to a goal or not, with every revision kept and a way to restore an old one. Agents may write them within limits of their own; a person is never limited. See Pages.
  • A status line. A sentence made from the board, such as “2 requests in flight, 1 needs you, 3 done this week, $12 of $60 spent”. A team can have its lead write it instead, which costs model calls and so is off until a person turns it on. That is not possible with an OpenClaw lead, or a Clawboo Native lead whose board access is read-only (the default) or off. See The status line.
  • Goals on the Board. Board cards and team chat task cards show the goal a task counts toward, its own or one it inherits from its request. The task drawer, New task and a request sent from team chat can pick a goal, and the Board’s filter can show one goal’s work, or only work toward no goal. See On the Board.
Behind a sign-in proxy
  • Roles. Set proxyIdentityRoles to enforce (or CLAWBOO_PROXY_IDENTITY_ROLES=enforce) together with proxyIdentity: required, and the role in each person’s identity token decides what they may do. An owner sees and does the work, answers approvals, and alone may answer one with “Allow always”. An operator sets up and runs the install but never answers an approval. A member sees and does the work, and answers approvals only when the token says "approve": true. A guest reaches only the teams the token lists, until its access_until. Give the operator role only to people you trust with the server. Roles are off by default. See Roles.
  • Only signed requests act as a person. Set personCheck to verified (or CLAWBOO_PERSON_CHECK=verified) with the proxy identity lane on, and answering approvals, Retry, pausing and the other person writes need a token your proxy signed: the dashboard page’s own marks no longer count. A card moved without one still moves, but as no one, so it stops no run. The same setting opens Clawboo’s Gateway WebSocket, and every route of an agent’s screen, only for such a person. It is off by default. See Only signed requests act as a person.
  • Take over only with the proxy’s say-so. With proxyIdentityControlClaim: required (or CLAWBOO_PROXY_IDENTITY_CONTROL_CLAIM=required), taking over an agent’s screen, and every batch of input after, needs a control claim naming the agent in a request-bound token. A batch without it hands the screen back to the agent. It is off by default. See Taking over a screen.
  • A key of its own for a program beside Clawboo. A program running beside Clawboo, such as one that watches for approvals, can make its own background reads with a key you set in proxyIdentityReaderPublicKey (or CLAWBOO_PROXY_IDENTITY_READER_PUBLIC_KEY). Its token reaches a short fixed list of reads and nothing else, and it never acts as a person. See A program beside Clawboo.
  • A used token stays used across a restart. The ids of the request-bound identity tokens Clawboo accepted are kept in its database until the token could no longer be accepted, so a token used before a restart is no longer accepted once more after it. When the database will not take a record, such as on a full disk, the token is checked against memory alone, as before, and the server logs an error.
The dashboard
  • A new frame. An account bar runs along the bottom of the sidebar, with the person a sign-in proxy verified (or “You” on a local install), the theme switch and the Settings gear, which moved there from the agent column. The team rail is 72 px wide with larger icons, and its add-team button sits at the end of the team icons and scrolls with them.
  • A near-black dark mode. Dark mode moves to a neutral near-black palette on every screen. The brand red, mint and amber stay, and light mode is unchanged.
  • A slim top bar in an app’s own window. An app that shows the dashboard in its own window can set data-clawboo-embed-titlebar on the page for a slim top bar: the sidebar’s hide button, back and forward through the dashboard’s own pages, the team or page name, Needs you and help. In a browser tab, nothing about the top of the page changes. See Ask for the slim top bar.
Runtimes
  • Connect this page to OpenClaw. A dashboard page that was already open when something else connected Clawboo to the OpenClaw Gateway, such as a routine for an OpenClaw agent or OpenClaw opened in another tab, can now connect itself without a reload. OpenClaw’s row in Settings > Runtimes reads Connected and offers Connect this page beside Manage while the page has no Gateway connection of its own, which OpenClaw 1:1 chats need. See Manage OpenClaw.
  • Hermes gets its planned stop. When Clawboo stops, it sends the Hermes gateway SIGINT and gives it up to CLAWBOO_HERMES_STOP_GRACE_MS (5 seconds by default, at most 8) to exit before it kills it and its processes. It used to get SIGTERM, which Hermes takes for a crash. clawboo stop now waits up to 15 seconds after its SIGTERM (it was 4) before it kills the server. See The Hermes gateway.
  • A runtime follows a moved Node.js. When the Node.js a runtime Clawboo installed ran on is gone, such as when an app that runs Clawboo on a Node.js of its own is moved or renamed, the runtime now moves to the Node.js Clawboo runs on, when that is the same version or a later release of the same major version on the same kind of processor. A runtime installed by 0.8.0 or older recorded no processor, so for it the processor is not checked: install it again if Clawboo now runs on a Node.js for another kind of processor. On any other Node.js it still reads as not installed until you install it again. See Which copy Clawboo uses.
  • Homebrew upgrades no longer change a runtime’s major version. A runtime installed on Homebrew’s node formula now names that Node.js by its versioned folder, so a brew upgrade to a new major version no longer moves it. A runtime installed this way before keeps the link: install it again after such an upgrade.
  • No Python bytecode from npm. Connectors started through npm, and the npm that installs a runtime into Clawboo’s runtimes folder, now start with PYTHONDONTWRITEBYTECODE=1, so a native part’s build no longer writes Python __pycache__ folders into the npm’s folder.
If you script the API
  • Home routes. GET /api/home reads a team’s Home, or the all-teams view, in one read, and the Home API has routes for goals, KPIs, goal updates, pages and Home settings. POST /api/board, a person’s board patch and a person’s team chat message take goalId. See Home API.
  • Home tools for agents. New Tasks MCP tools: list_goals, get_goal, list_kpis, list_pages and read_page to read, and, when their session is bound to them, report_kpi, suggest_kpi, write_page, draft_goal_update and, for the lead, write_home_summary. No tool lets an agent publish an update, set a goal’s health, define a KPI or change a target. See Home tools.
  • A start marker. The server writes api-port.json next to api-port.txt, naming its process and version, so a reader can tell a running server from the record of one that crashed. The clawboo command checks it before it trusts the recorded port. api-port.txt is unchanged. See Start marker.
  • Is the server up. GET /api/system/live (and HEAD) answers { "ok": true } with no access token and no identity token, and says nothing else about the install. See The liveness probe.
  • GET /api/access/identity now says whether roles are enforced and what the person may do, and answers personCheck, controlClaim and readerKey. New capability flags: auth.roles, auth.personCheck, auth.readerKey, computer.controlClaim, system.live, ui.embedTitlebar and home.goals. See Feature flags.
  • Four new event kinds record agent KPI reports, update drafts, publishing and page revisions. See Home events.
Fixes
  • A SIGTERM sent while the server was still starting could be taken by one part of the server, so the server kept running and its later stop left the OpenClaw connection open. A SIGINT or SIGTERM sent during the start is now kept, as the shutdown message already was, and runs once the server has started.
  • On a start that moves work folders, a stop sent during the start no longer lets the routines resume, and possibly start a run, in the moment before the stop.
  • On macOS and Linux, process start times are now read in UTC, so a process that started in the hour a daylight saving change repeats is no longer taken for another one.
  • A task that waits on or depends on a task of another team no longer copies that task’s title into its own comments, its waiting reason or its team’s room.
2026-11-15
0.8.0 lets an app that runs Clawboo stop it cleanly and open the dashboard from its own notifications. Clawboo now connects to the OpenClaw Gateway only once something needs it, so a Gateway that starts on its first connection can stay asleep. On Linux, each agent can run as an operating system user of its own.Run Clawboo from your own app
  • Stop it over the IPC channel. An app that runs Clawboo as its own child process can send { "type": "shutdown" } on the channel, and the server stops the way it does on SIGTERM: it ends the agent runs, closes the database and exits with code 0. This works on Windows too, where a signal ends the process at once. With CLAWBOO_STOP_ON_DISCONNECT=1, the server also stops when that channel closes, so it stops with an app that quits or crashes. A server the clawboo command starts keeps running after the command returns, as before. See Start it from your own app.
  • Open the dashboard from a notification. An app that shows the dashboard in its own window can open what one of its notifications is about, through onOpenTarget on the object it gives the page: the task’s drawer, or Needs you with the approval’s card brought into view and outlined while it has focus. Before, the app could only bring its window forward. See Open the dashboard from a notification.
OpenClaw
  • The Gateway is left alone until something needs it. Clawboo connects to the OpenClaw Gateway once there is an OpenClaw agent (or a delete of one still owed to the Gateway), once you open OpenClaw in Settings > Runtimes or start the Gateway from Clawboo, or once a routine for an OpenClaw agent falls due. An install with OpenClaw agents connects at every start, as before. Until then neither the server, the dashboard nor clawboo connects to the Gateway or probes its port, so a Gateway that a service manager starts on its first connection, such as a systemd socket unit on a small Linux machine, stays asleep instead of starting at every boot. See How Clawboo connects.
  • Not in use, rather than unreachable. OpenClaw’s row reads Not in use with Connect while nothing uses the Gateway. The welcome screen offers Open OpenClaw to bring in the agents of an OpenClaw already set up, Capabilities shows no warning for it, System Health passes its Gateway check as “not in use yet”, and the dashboard loads with no offline banner.
  • clawboo no longer prints whether a Gateway answers on port 18789 when it starts.
A user per agent, on Linux
  • Each agent can run as an operating system user of its own. Set CLAWBOO_AGENT_USERS to a pool of users you create once (names, or a uid range such as uid:20001-20008), CLAWBOO_AGENT_GROUPS to a pool of team groups, and CLAWBOO_WORK_DIR to a folder of your own. It is used instead of CLAWBOO_AGENT_USER, never with it, and is off unless you set it. One agent then cannot read another’s sign-ins, sessions or cookies, reach its display or signal its processes. Teammates read and change each other’s files in their team’s folder, and agents on other teams cannot reach it. See A user per agent.
  • An agent keeps its user for life. An agent gets the next free user when it is created, and each team gets a group the same way. A deleted agent’s processes stop at once, and its user is wiped before another agent gets it. With no free user, a new agent is still created and its runs say why they cannot start; it never falls back to a shared user.
  • Set it up and check it. clawboo agent-user plan --pool <n> --teams <m> prints the setup, with the one sudoers rule a pool needs, and clawboo agent-user check checks a pool too. The startup check now also proves, from inside each agent user’s own host, that no agent user can list another’s home or signal another’s process, and refuses a pool whose users share a uid, a primary group or a home.
  • Not available yet with a user per agent: code tasks that work in a checkout of a repository (refused with the reason), Hermes (it does not start), OpenClaw (not covered, as before), connectors with local reach shared by every agent (they do not start; each agent’s own browser does), picking an agent for a Codex or Hermes sign-in on the Runtimes page (give the runtime an API key), a container launcher for a pool, macOS, and copying older agent state in. See Not available yet with a user per agent.
If you script the API
  • For a Gateway nothing uses, GET /api/agents is not stale, GET /api/agents/registry/health and GET /api/system/status report inUse: false, and GET /api/system/idle reports the Gateway’s cron jobs off with the reason not_in_use. The registry health also reports configured. See Agents API.
  • New route: POST /api/agents/registry/connect, which the dashboard sends when OpenClaw’s row opens.
  • With a Gateway set up that nothing used yet, creating an OpenClaw agent or a sync without a body now waits up to about 5 seconds for the connection.
Fixes
  • A browser connection to the Gateway that fails no longer keeps retrying unseen.
  • A reconnect asked for while a Gateway connection was still being made no longer leaves the server unable to connect until it restarts.
2026-11-14
0.7.0 makes the Board the place where work is tracked and decided, and gives each agent a computer you can watch. Tasks get readable ids such as MAR-12, a request you send in chat becomes one main task with its sub-tasks under it, a plan can wait for your OK, and a new Inbox shows what needs you and what changed. You can pause an agent, a team or a whole task tree, and Boo Zero can hand work to a team from its own chat. Runtimes now install into a folder Clawboo owns, with no sudo, and Hermes runs as the real Hermes through its own gateway. Agents can run as their own operating system user, each task gets a folder of its own, you can watch each agent’s browser live, and agents can sign in to websites with a login they are never handed.The Board
  • Readable task ids. A team’s tasks read like MAR-12: letters from the team’s name, and a number. They show on cards, in the drawer and on task cards in chat, and Find and the Tasks tools take them. You can change the letters in the team’s settings until its first task. Existing tasks get theirs on the first start. See Task ids.
  • One task per create. A create sent again from the New task dialog, the Tasks tools or a team chat hand-off returns the same task instead of making a second one, and an agent that hands off the same open work twice under the same parent within 48 hours gets the first task back. An agent cannot hand work back to an agent that is waiting on it, and the hand-off caps now hold in team chat too.
  • Your request is one card. When a team’s lead hands out work for something you asked in chat, your message becomes a main task with every hand-off and plan step nested under it, on the board and in the chat. It waits for its sub-tasks and moves to Done after the lead’s summary. To add to a request, comment on its main task; a new message in the chat is a new request. See Your requests and their sub-tasks.
  • A task that hands out work waits for it. In team chat, its card reads “Waiting on 3 sub-tasks”. When they close, its agent gets one more turn with all their results. A failed sub-task brings the parent to Needs you instead of letting it close as done. Agents can say how a task went with the new report_result tool, which never gets a task past a failing check.
  • Plans. An agent can write a plan as a document on its task, with every revision on the drawer’s new Plan tab. Turn on Ask me before splitting work for a team, and a plan waits in Needs you until you accept it. It is off by default. See Decide on a plan.
  • Every waiting task says what it waits for. An agent that blocks a task must give a reason, and Needs you shows it. A task can be assigned to a person, and then no agent takes it. An agent can ask you a question (a choice, a yes or no, or a list of tasks to accept), which parks the task until you answer. You can reopen a finished task. On a local install, “a person” is one built-in person.
  • Agents start when something happens on their task. Your comment, or a task it waits on finishing, is saved so a restart cannot lose it, and starts the right run once. Agents can start each other with a comment once you allow it: Needs you asks once per job, and after that a job allows 12 wakes an hour, and two agents answering each other stop after 4. A comment from you always wakes the agent. See What wakes an agent.
  • A run’s ending is told apart. An empty reply gets one more turn, then waits in Needs you as No result instead of landing in Done. A rate limit, overload or network error goes back to To do and tries again twice. A budget cap or the circuit breaker stopping a run has a status of its own, so neither reads as your Stop.
  • A job cannot run away. Clawboo counts the runs a job starts by itself. At 100, nothing more starts by itself, and the job’s top task shows in Needs you with Carry on and Stop. Your own messages, Retry, replies and Accept never count. See When work restarts by itself.
  • Labels, and a fair order. Teams have labels you can set on tasks and filter the board by. Ready tasks go in this order: tasks that ran and came back, then priority, then oldest first. A task that keeps waiting moves up one priority every 4 hours, up to High. Agents, and scripts other than the dashboard, can set a priority of High at most: Urgent is a person’s call.
Inbox and pause
  • Inbox. The sidebar has an Inbox with Unread, Needs you, Mine and All, and a badge that counts what needs you and what changed. What you do yourself does not make a task unread for you. Read state is kept on the server, so it is the same in every browser, and behind a sign-in proxy each person has their own. See See it all in the Inbox.
  • Pause an agent, a team or a task tree. Use Pause in an agent’s or a team chat’s header, in a task’s drawer, or from a right-click on the team rail or the Org chart. A pause holds until you resume it, through a restart. Its runs stop within seconds, and nothing it covers starts: no task, no Routine and no wake. A routine that falls due during a pause runs once on resume. A one-to-one chat with an OpenClaw agent runs in the Gateway, so a pause does not cover it. See Pause work.
  • Archived teams stop. An archived team’s agents stop working, and nothing wakes them until you bring the team back. Before, work left on its board kept starting.
  • A person goes first. When every run slot is taken, a person’s chat turn goes first, then tasks by priority, so an Urgent task is no longer stuck behind everything queued before it.
Boo Zero
  • Hand work to a team from Boo Zero’s chat. Start a message with @Team, or say which team in plain words, and each hand-off becomes a task on that team’s board, with its card in both chats. The result comes back to Boo Zero’s chat, even across a restart. This works when Boo Zero runs on Clawboo Native, Claude Code, Codex or Hermes; an OpenClaw Boo Zero cannot hand off from its chat, and says so. See Give a team work from this chat.
  • Boo Zero asks before picking a team you did not name. Work starts at once only on a team your message names. A hand-off to any other team waits on a card in the chat, and nothing starts until you press Accept.
  • Words that are not yours are marked. What an @Team message pulls in, Boo Zero’s saved notes, and results kept since a Stop reach Boo Zero quoted, under who wrote them. On Clawboo Native, Claude Code, Codex and Hermes, a one-to-one chat gets only get_task, list_tasks and add_comment on the board.
Agent computers
  • Watch each agent’s browser live. The Browser card on an agent’s page shows its browser live, and falls back to the last screenshot when no browser is open. On macOS and Windows, and on Linux without Xvnc, the view shows the browser about once a second, for watching only. See Watch a screen and take over.
  • A screen for each agent on Linux. With TigerVNC’s Xvnc on the server, each agent’s browser opens on a virtual screen of its own. Take over sends your mouse and keys to it while the agent’s browser tools wait, until you Hand back. The agent’s other programs share that screen, so type a password there only for an agent you trust. The screen listens on no network port. All agents share one agent user, so this keeps displays from other users on the machine, not agents from each other.
  • Install the desktop tools. On Linux, Install the desktop tools on an agent’s Screen view installs TigerVNC’s Xvnc (x86_64 only) and Cua Driver into Clawboo’s runtimes folder, each checked against a SHA-256 Clawboo pins. Clawboo’s copy of Xvnc needs some system libraries that Clawboo does not install, and the Screen view names any that is missing. Cua Driver, a new connector, lets an agent use the apps on its own screen. It runs only when agents run as their own user, and only for the agents you give it to. See Connectors.
  • Share a screen with a link. A view link lets someone with no way into your Clawboo watch one agent’s screen. A control link, for example for a person who must type their own password into a site, is offered only when agents run as their own user. A view link expires after 8 hours and a control link after 1 hour by default. You can revoke either, and a restart revokes them all.
  • The Computer tab. The tab on an agent’s page that showed its files is now Computer, with Screen and Files. It lists the agent’s task folders beside its worktrees.
  • A folder for each task. A task run with no git worktree now starts in a folder of its own, ~/clawboo-work/teams/<team id>/tasks/<task id> by default, instead of the team’s shared folder, so a file stays with the task that made it. The task drawer shows it under Files. The folder sets where the run starts; it is not a sandbox. See Task folders.
  • Sign in to websites without handing over the login. Save a login under Settings > Website logins, or connect 1Password with a service account token. On a sign-in page, an agent calls the new fill_login tool, and Clawboo fills the login into the agent’s own browser. The agent is told only filled, no saved login, needs you or refused. A login fills only on a page with exactly its saved scheme, host and port. Once a field is filled, the agent can still read it in its own page. A run uses only the logins of the person who started it or asked for its task, and a routine’s run uses none. fill_login is offered to Clawboo Native, Claude Code, Codex and Hermes agents while the Playwright or Chrome DevTools connector is connected. See Website logins.
  • Each agent’s browser profile has Chrome’s password manager turned off, so no password typed or filled into it is saved there.
Runtimes
  • Runtimes install into Clawboo’s own folder. OpenClaw, Claude Code and Codex install into ~/clawboo-runtimes instead of npm’s global folder (<CLAWBOO_HOME>-runtimes when CLAWBOO_HOME is set; CLAWBOO_RUNTIMES_DIR sets another). An install no longer needs sudo, no longer fails with a permission error on a system-wide Node, and is not lost when nvm switches Node. An update installs beside the current copy and switches only when it succeeds. A runtime you installed yourself keeps working when Clawboo has none of its own. See The runtimes folder.
  • OpenClaw checks Node first. On a Node outside OpenClaw’s range (>=24.22.1 <25 || >=26.12.1), the install says so and installs nothing, where npm used to install an older OpenClaw without a word. OpenClaw’s setup also writes its tool sandbox as off, so OpenClaw’s tools run on the same computer as your other agents; a sandbox you set up yourself is kept.
  • Hermes runs as the real Hermes, through its own gateway. Installing Hermes from Settings > Runtimes fetches release 0.23.1 from GitHub, with the uv and Python builds that release pins, each checked against a pinned hash: no Python of your own, and no pip or pipx. A Hermes installed earlier with pip or pipx reads Update needed until you install Clawboo’s own. Clawboo runs one Hermes gateway on 127.0.0.1 and drives every Hermes agent through it as a Hermes profile, so each agent’s memory, skills and sessions carry over. Hermes runs now stream, take a comment into the running turn, report their tokens, and ask before the commands Hermes’s own policy flags. Hermes’s chat channels stay off. See The Hermes gateway.
  • Hermes’s own jobs show in Routines. Jobs a Hermes agent schedules in Hermes show in Routines under Hermes’s own jobs, where you can disable, enable, run or delete them. Team work stays a Routine. CLAWBOO_HERMES_TRANSPORT is gone.
  • Send a provider through your own gateway. Set a provider’s base URL, such as OPENROUTER_BASE_URL or ANTHROPIC_BASE_URL, to route its traffic through a gateway such as LiteLLM or a company proxy. Clawboo Native’s runs, the key check when you save a key, and the model list all use it. An administrator can set it in the operator settings file too. See Provider base URLs.
  • A key from the environment shows as connected. A provider key in the environment Clawboo starts with, such as ANTHROPIC_API_KEY, now reads as connected on the Providers page, and the row says it comes from the environment. Setting up OpenClaw still reuses only a saved key, so a key from the environment is never written to disk.
Agents on their own user, and the work folder
  • Run agents as their own operating system user. Set CLAWBOO_AGENT_USER to a user made for them, and Claude Code, Codex and Hermes runs, Clawboo Native’s approved commands and file tools, a task check’s verify command, git in task checkouts, connectors with local reach and each agent’s browser run as that user. That user cannot read Clawboo’s database, the vault, its key, settings or the access token. Before it accepts a run, Clawboo checks this from inside that account. When the check fails, Clawboo refuses to start and says what to change. clawboo agent-user plan prints the setup commands for Linux or macOS, and clawboo agent-user check runs the check. It is off unless you set it and is not available on Windows. It does not limit the network, keep agents apart from each other, or cover OpenClaw agents. Clawboo stops every process the agent user runs each time Clawboo starts or stops, so give each install an agent user of its own. See Run agents as their own user.
  • Your sign-ins stay yours. With the setting on, your own Codex and Hermes sign-ins are not shared with the agents: sign them in from the Runtimes page, which runs as the agent user, or give the runtime an API key. Task checkouts come from a copy of each repository that the agent user owns, so an agent can no longer plant a hook in your repository. Connectors that hold a saved credential keep running as you, so agents cannot read their tokens.
  • Agent work moves out of the data folder. Team folders, task worktrees and review checkouts go in a work folder, ~/clawboo-work by default (CLAWBOO_WORK_DIR sets another), so ../.. from a team folder no longer leads to the database or the saved keys. Agents that run as you can still open ~/.clawboo by its full path: this removes the short route, not the access. The first start moves older work there by renaming it, and leaves a link at each old path. You can go back to 0.6.x after the move, though its team chat work checks then fail until you upgrade again; do not go below 0.6.0. See The work folder.
  • Clawboo no longer follows a link, or a hard link, that an agent puts at a task file’s name such as AGENT_HANDOFF.json or init.sh.
If you script the API
  • A person’s writes go through one check. Board writes now record who made them. Moving a card, Retry, Assign, answering an approval, Stop on a chat, and the other writes that decide for you, such as a team’s rules and members, budgets, Routines and their time zone, and connector grants, take a person’s authority only from the dashboard’s own page or from a person a sign-in proxy vouched for. Any other caller gets 403 person_only, except that a card move from it still goes through, without a person’s authority. A script acting for you can send Sec-Fetch-Site: same-origin and Sec-Fetch-Mode: cors. Chat messages from other callers still go through, marked Local API call. On your own computer this is a speed bump, not a security boundary. See The person check.
  • Copy signed MCP URLs for external agents again. An attach signature now names the server it is for, so a Memory URL’s signature no longer opens the Tasks tools. A signed URL you copied from /api/mcp/config is served as an unverified caller until you copy it again. Clawboo writes new URLs for its own runs each run.
  • An app that starts Clawboo can name its person. It hands Clawboo its public key in CLAWBOO_APP_LINK_PUBLIC_KEY and signs a short token for each request, so a message sent through the app counts as that person’s and can start work. Such a person can allow an approval once or deny it, but never allow it always.
  • Know when nobody is using Clawboo. GET /api/system/idle says whether a person is here, whether an agent run is going, and when the next routine, OpenClaw cron job or Hermes job is due, so a computer or service that stops Clawboo when idle can start it again just before. See Stop it when nobody is using it.
Fixes
  • The server answers sooner after it starts. Its startup database check is now SQLite’s quick check, in a background thread; the full check runs once a day and on Re-run probe in System Health.
  • A due routine on an OpenClaw agent waits up to 5 minutes for the Gateway to connect instead of failing, so a routine that came due while Clawboo was stopped or the computer was asleep still runs.
  • A one-to-one chat no longer resumes a session that outgrew the model’s context window. On Claude Code, Codex and Hermes the next message starts fresh, with one notice in the chat; Clawboo Native first trims its oldest turns when that makes room. Retry on a board task whose Claude Code run overflowed now starts a fresh session instead of overflowing again.
  • A team chat work check waits while a teammate is still working in the team folder, for up to 15 minutes, so it no longer checks a change halfway through. The card reads Waiting to check. A busy folder no longer lets finished work reach Done unchecked. The wait does not keep teammates’ changes apart.
  • An agent’s reply goes on its task when the work check starts, so a restart during the check no longer loses it. A crash while the task runner checks work no longer leaves the card stuck on Verifying.
  • Deleting a team stops its runs and cancels its open requests and the work under them.
  • Costs are charged where they belong: the reviewer asks the budget before it spends, and a one-to-one chat that reads a team’s results is charged to that team. The audit log records budget edits, and writes the 80% and 100% warnings for team chat, one-to-one chat and peer-exchange spend.
  • Task comment and team chat events keep their author fields readable in the event log, instead of [REDACTED].
2026-11-04
0.6.0 gives a team chat rooms, lets you steer a task while an agent works on it, and rebuilds the workspace around a team. The Board opens on a calmer Kanban view. Clawboo Native runs on a local Ollama with no key, and a security gap for installs behind a reverse proxy is closed. The dashboard now locks itself with a per-install access token, installs as a web app on a phone, and clawboo stop and updates wait for agents that are still working.Security
  • The dashboard is locked to your computer by default. With STUDIO_ACCESS_TOKEN unset, the server makes an access token on its first start and keeps it in ~/.clawboo/access-token, readable only by you, so other user accounts on the computer can no longer use the API. A program running as you still can, as it can read your database. clawboo and clawboo restart open the browser with a one-time sign-in link that works once, within two minutes, and print only the plain address, now http://127.0.0.1:<port> rather than localhost. A chat app’s link preview does not spend a link; a phone on a plain-HTTP network address gets a Sign in button instead. The CLI sends the token only to the port the server recorded. The new clawboo open signs in a browser for a running server, and clawboo open --print prints a link for a computer with no browser. A browser that is not signed in shows “This dashboard is locked to your computer”. Open on another device, at the bottom of the sidebar, makes a one-time link, with a QR code for a phone that can reach the server. STUDIO_ACCESS_TOKEN= set blank still turns the gate off, and pnpm dev runs that way. A script that calls the API needs the cookie: see Authentication, and Security for what the token does and does not protect.
  • The access token covers the MCP servers behind a proxy. On 0.5.0 and older, any request from 127.0.0.1 reached /api/mcp/ without the access cookie, so a reverse proxy on the same machine exposed the MCP servers and let anyone ask /api/mcp/config for a signed attach URL. Now a request without the cookie reaches an MCP server only with an attach token, only on a direct connection to 127.0.0.1, and never through a proxy. /api/mcp/config always needs the cookie. Until you upgrade, deny all of /api/mcp/ at the proxy: see Security model.
  • Re-copy attach URLs you set up by hand. If you attached an agent by URL on an install with an access token, copy the URL again from GET /api/mcp/config (sending the access cookie). The old URL carries no token and gets a 401.
  • A token the access check cannot use stops the server. A STUDIO_ACCESS_TOKEN with a character other than letters, digits, ., _, ~ and -, such as openssl rand -base64 output, used to turn the check off. Now the server logs SECURITY: refusing to start and exits; make a new token with openssl rand -hex 24. Files that hold keys or tokens, including clawboo backup files, are now readable only by you.
  • Claude Code agents leave your own Claude Code setup out. They no longer load your ~/.claude settings, plugins, hooks, MCP servers, CLAUDE.md and rules, or Claude Code’s auto memory. A repository’s own CLAUDE.md and .claude/settings.json still apply. Auth you set in ~/.claude/settings.json, such as an env block or an apiKeyHelper, no longer reaches agents: set those variables in the environment you start Clawboo with. See Claude Code.
  • Connector access set to ask every time always asks. A remembered Always used to skip the prompt even under approvalPolicy: always. Now the call asks, and the card offers no way to remember it.
  • Risky hand-offs are never remembered. An allow_always answer sent through the API used to stop that leader’s risky hand-offs from asking again, with no end date and no way to undo it in the dashboard. Now it allows that one hand-off.
  • Attach URLs name real teams, and their secret can be replaced. GET /api/mcp/config refuses a scopeTeamId that names no team, and a Clawboo agent signed for a team it is not on. POST /api/mcp/attach-secret/rotate replaces the secret behind every attach URL without a restart. See MCP servers.
  • TeamChat takes no author from tool arguments over HTTP. A session without a signed room-and-author binding gets no TeamChat tools.
  • create_subtask checks the team. A parent task from another team is refused.
  • Images in an agent’s text no longer load on their own. In chat, a task’s report and its comments, an image used to load from its address as soon as it was shown, so an agent steered by something it read could send data out in that address. It now shows as a link named after its alt text that opens only when you click it. An address that is not a web or same-site one is only named.
  • A link in a worktree is never served. The task drawer’s preview checked a file and then opened it by name, so an agent could swap the file, or a folder above it, for a link to a file such as Clawboo’s saved keys, and the preview sent that file. The preview, an agent’s Workspace tab and the task drawer’s task files now refuse a link or a hard-linked file and check after opening that the file is still inside the worktree. A task file longer than 256 KB says so in the drawer.
  • API routes match letter case exactly, and with no access token set, the server logs a one-time SECURITY: warning when a request arrives through a reverse proxy.
  • The access token and the same-origin guard read a path the way the router does. Both now check every request the server routes to an /api/ route, however its path is written. A request whose target is a full URL instead of a path gets 400 (404 under a path prefix); browsers, reverse proxies and Clawboo’s own clients always send a path.
Rooms and chat
  • Rooms. Next to the team’s own chat, now called General, a team can have rooms, each with some of its members. A room has its own conversation, task cards and graph. Boo Zero leads every room, and the team’s rules, brief, memory and board stay team-wide. Create one with New room, and pin, rename or delete it from its menu. Deleting a room archives it. See Team chat.
  • Chat with Claude Code, Codex and Hermes agents. The chat on their page now runs on the Clawboo server, with no OpenClaw Gateway, and each agent keeps its conversation from one message to the next until you send /reset. These chats and Clawboo Native’s start in the agent’s team folder, where a Clawboo Native agent can now use its file tools and, with Let this Boo ask to run commands on, ask to run a command. See Agents.
  • Clear chat. Every chat’s menu can clear it. The messages are set aside rather than deleted, the agents start that conversation fresh (OpenClaw agents are sent /reset), and the chat stops showing task cards from before the clear. Those stay on the Board and in the Tasks card.
Steering
  • Steer a running task. Comment on a task from a team chat’s Tasks card, or from Tasks & Routines on an agent’s page, and the agent reads the comment during the run. Clawboo Native, Claude Code, Codex and OpenClaw take it at the run’s next step, and Hermes after the tool call in progress. A Hermes installed without its live mode pauses the run and resumes the same session with your comment instead; its row in Settings > Runtimes then reads Update available and offers Update Hermes, and the line beside Send says which way the run on a task takes it. The comment always lands in the agent’s mailbox as well, so a task that is not running yet still gets it. See Steering a running task.
  • Codex runs in codex app-server, which is what lets it be steered. An older Codex without it falls back to codex exec, where a comment waits for the next run.
  • A reply reaches the agent, and says where it went. A reply on a task now goes to the agent it belongs to whatever the task’s status, including a task moved back to To do. A reply to a task waiting on you, or waiting in To do, starts the agent on it again, so answering a blocker no longer needs a second click on Retry; replies within 15 seconds start one run. A reply to a task in review waits for the agent to read after the check. A finished task cannot run again, so the reply is not sent, and the note under it offers Send as a follow-up task, a new task for the same agent. The Board’s task drawer now has a reply box, and a note under each reply you send, there and in the chat’s Tasks card, says where it went. The Tasks card’s Send and retry button is gone, since Send does both. See Steer a running task.
The workspace
  • A sidebar you can resize and hide. Drag the agent column’s edge to resize it, or hide it with Cmd/Ctrl + \. The nav reads Home, Org chart (the new name for Atlas), Board, Connectors, Memory, Marketplace and Settings, and Cmd/Ctrl + 1 to 4 open Org chart, Board, Connectors and Memory. Fleet moves to Settings > Insights.
  • Cards beside every chat. An agent’s page shows Graph, Tasks & Routines, Browser, Activity and Details next to its chat. A team chat shows Browser, the team’s or room’s graph, Tasks and Activity.
  • Home. Each team has a Home page, the place its dashboards and its agents’ artifacts will live.
  • The Org chart turns. With the browser panel open, the radial Org chart turns around Boo Zero to bring the team you pick next to the panel.
Ollama and providers
  • Ollama with no key. Picking Ollama during setup now counts as a connected provider everywhere: agents deployed without a model pick run on it, Boo Zero is created on it, and team chat and one-to-one chat both work. Turn it on or off in Settings > Runtimes > Clawboo Native > Manage.
  • Your pulled models. The model lists show the chat models your Ollama has pulled, mark the ones that cannot use tools, and show ollama pull llama3.2 with Check again when none is pulled.
  • Your default provider wins. Agents deployed without a model pick run on the provider you chose as your default, instead of the first one with a key.
  • OpenClaw on Ollama. When Ollama is your provider, Set up OpenClaw configures OpenClaw on it instead of asking for a key, and stops with the ollama pull command when the model is missing. Setup also connects to the Gateway port in your OpenClaw config.
The Board
  • Kanban first. The Board opens on Kanban (the view called All before, key K), every task by status with Needs you first. Today, Week and Month stay as the calendar views, and a view you already picked is kept. On Kanban, cards say how long ago they were added or finished (“4m ago”, “Yesterday”, “3d ago”); the calendar views keep clock times.
  • One header and one toolbar. The header holds the title, the team filter, Star and New task (key C). Refresh and the subtitle are gone; a spinner shows only when a background read is slow. The toolbar holds Find (/ or Cmd/Ctrl+K), Filter (agents, tasks with no agent, what is running now), Display (routine runs, the Cancelled column, cost and time on cards), an agents-working chip and the view menu. “Since you last looked” is a chip that shows only when something changed.
  • Calmer cards. Cards and lanes are neutral, with color only in each column’s status mark and amber for Needs you. Cards drop the runtime mark, the corner clip and the hover lift, and counts no longer roll.
  • New task in one sentence. Write what should get done and press Enter. Clawboo suggests who should do it by matching your words to each agent’s name and the role in its SOUL.md, else the agent you last gave a task to, then the team lead, and says why. Picking an agent also picks its team. More options opens the full form with Team, Agent and Details.
  • Needs you cards lead with one action. Each card says why it waits and offers Retry, Assign, Review or Open, with the other choices, including Dismiss, in a More menu.
  • The task drawer reads like a document. The agent’s report once, as formatted text, then the brief, the properties, verification, the workspace, run history, comments and activity. Each section folds from its heading, empty ones are not shown, and J and K step through the column. See Board.
Budgets
  • Pick a budget by name. In Settings > Insights > Governance, choose agent, team or mission, then the agent, team or board task from a list you can search. You no longer look up an id first, and each budget shows its name. The form no longer offers tenant, which nothing charges; a tenant budget you already have still shows.
  • Runs that start together no longer overshoot a hard cap together. Each run holds its estimated cost on its budgets from the moment it is let start until it ends, and is checked again right before it calls the model. On a nearly spent hard cap, a run that would go beside others starts only when all of them fit; otherwise it is refused with a notice saying the rest of the cap is set aside. A run going alone, and every warn-only budget, behave as before. Holds are estimates, so runs can still cost more than they held. See Runs that go at the same time.
  • One-to-one chat shows in Tokens Used. Every one-to-one turn with a Clawboo Native, Claude Code, Codex or Hermes agent writes the same cost record a team-chat run does, from the runtime’s own figures or estimated from the text, whether or not the agent has a budget.
Fixes
  • Creating and deleting agents works without an OpenClaw Gateway, and a new agent made on the Org chart always joins a team.
  • An OpenClaw agent deleted while its Gateway is down stays deleted.
  • Approving this device during OpenClaw setup shows what went wrong, with Retry, instead of sticking on “Approved. Reconnecting…”.
  • Updating from the dashboard keeps working after the first update, a failed update shows npm’s own reason, and Clawboo runs openclaw doctor --fix --non-interactive once when a newer OpenClaw needs it before the Gateway will start.
  • The Org chart’s Radial layout spaces each team’s Boos apart, also with “reduce motion” turned on.
  • While one of a routine’s runs is still going, the Board’s calendar marks the right next run.
  • Loading the dashboard no longer counts against the API’s request limit.
  • A Boo no longer loses its body when another copy of it is hidden.
  • An agent can no longer put a card back after you moved it: its own status changes are refused, with a message telling it not to try again, until the task runs again.
  • The Kanban lanes run unbroken to the bottom, and dropdown menus are as wide as their longest option.
  • Team chat runs, and task runs that only read (routines made in the dashboard included), start in a folder of their team, ~/.clawboo/workspaces/<team id>, instead of the folder that holds Clawboo’s database and keys. The folder only sets where Claude Code, Codex and Hermes start: they keep full access to your files. Clawboo Native agents get their file tools there, kept inside the folder. An agent’s Workspace tab shows the folder, and deleting a team deletes it with its files. See What agents can do.
  • A model your own Ollama serves is recorded at $0 and reads free, a Claude Code or Codex run whose final cost report crosses a hard cap keeps its result (the next run is refused instead), every run that ended without a cost report is estimated, and Tokens Used lists every runtime. See Budgets and spend.
  • A paused hard cap now refuses new team-chat and one-to-one chat turns before they start, with a notice in the chat saying which budget is paused; a refused hand-off waits in Needs you with the reason. Saving a budget again keeps its mode unless you pick another, and the Governance form shows the mode of a budget you already have. See What a hard cap stops.
  • With an access token set, the clawboo command uses the STUDIO_ACCESS_TOKEN from the shell you run it in. It now sees a token-protected server it just started, instead of reporting Dashboard is taking too long to start. while the server runs, and clawboo stop and clawboo restart work on a server whose token that shell has. See Token-gated installs.
  • The reviewer that checks large or handed-off task-runner work is charged to the same agent, mission and team budgets as the work, and shows in Tokens Used. Completing a worktree through the API also works on a task back in To do. See How work is checked.
  • A checked task-runner agent is told up front, in its instructions and its worktree’s TASK.md, to set VERIFY_CMD, so the first check no longer always fails on the placeholder. A task whose title or description names a risky action, such as deploy, delete or production, always gets the reviewer.
  • A runtime’s Disconnect keeps the keys other runtimes also read and deletes only the ones that runtime alone reads; its dialog names each one. Claude Code and Hermes can connect again with a kept key through Use saved key. See Disconnect a key or a runtime.
  • A task that was in review when Clawboo restarted waits in Needs you with a comment, instead of going back to To do and running again.
  • A task that runs again no longer loses uncommitted work when its worktree is rebuilt. A checkout still on the task branch with nothing uncommitted is reused as it is. Any other folder that holds uncommitted changes, files git ignores such as .env, a commit no branch holds, a separate git repository, or contents git cannot read is left as it is, and the task waits in Needs you with a comment naming it; when git does not answer the check, the run does not start and nothing changes. Startup cleanup skips such folders too. See Worktrees and handoff.
  • The in-app Hermes install picks a Python from 3.11 to 3.13, so a newer 3.14 no longer wins over an installed 3.12, and a Hermes agent on your ChatGPT plan stays on it: its page shows its model instead of offering OpenRouter models that would move it off the plan. See Hermes.
  • A Marketplace team whose suggested runtime is not available starts its members on the first runtime that is connected, such as Codex on a ChatGPT-only setup, instead of on a Clawboo Native with no provider. Pressing Deploy again after “No agents could be created” now deploys the team, and the failed try leaves no empty team behind.
  • POST /api/team-chat/exchange runs its turns again. It used to cancel every exchange at its first turn.
  • Hermes agents get Clawboo’s Tasks, Memory, Tools and TeamChat servers. Clawboo used to list them in a file Hermes never reads, so a Hermes agent could not claim tasks, post in team chat, use memory or call connectors. They now go in the mcp_servers block of the agent’s config.yaml, next to any servers of your own. A long Hermes run also reports that it is alive once a minute, so a delegated Hermes task that works for more than eight minutes is no longer stopped. See Hermes.
  • A Claude Code run that hits an API error, such as a bad key, an expired sign-in or a billing limit, fails with the error and its cause instead of landing its task as done with the error text as the result.
  • Google agents on Clawboo Native start on gemini-3.6-flash; Google shut down gemini-2.0-flash on 2026-06-01. Multi-step tool turns on Gemini 3 keep their thought signatures. Anthropic runs on Clawboo Native, direct or through OpenRouter, use prompt caching, and cached tokens are priced at the cache rates. See Clawboo Native.
  • Connecting OpenClaw with a Google key sets google/gemini-3.6-flash as the default model, and the OpenClaw model picker lists Gemini 3.6 Flash in place of Gemini 2.0 Flash. A default an earlier version set is kept: change it under Manage > Default model. See OpenClaw.
  • Cleaning up connector processes a crashed server left running no longer risks stopping an unrelated process. Inside a container, a restart kept the host’s boot time, so a process id recorded before the restart could name another process. Each record now names its PID namespace and when that namespace’s first process started, and a process whose start time cannot be read is left running and checked again at the next start. Agent processes recorded before a container restart are never stopped either.
  • A team can have a verify command that every check runs, set under Work check in the team’s settings or with PUT /api/team-verify-command/:teamId, and each check records where its command came from: the task drawer flags one the agent wrote itself. A review that failed or returned nothing reads Review did not complete or Review returned nothing instead of Reviewed by. See How work is checked.
  • Claude Code reads as installed exactly when the Claude Agent SDK and its own claude binary sit next to Clawboo, because every run loads the SDK and starts that binary. A claude CLI on its own no longer reads as ready while its runs fail, and Install fetches the SDK (@anthropic-ai/claude-agent-sdk@0.3) instead of the CLI; from npx or a source checkout it says what to run instead.
  • Routines run in a time zone you choose, one for every routine on the install and independent of the server process. The routine dialog names it under Runs, with Change to pick another, and a change moves every routine’s next run to the new zone. It starts as the server’s zone, so nothing changes until you pick one. A server in a container, which runs in UTC unless TZ is set, can now run a 9:00 routine at 9:00 your time. A once@ time typed in the dialog needs a Z or an offset.
  • Connectors and the shell commands you approve keep npm’s prefix, cache and update-check settings (never an npm credential), so an approved npm install -g works on a server whose npm prefix points at a folder of its own. Connectors also start with npm’s update check off.
Running Clawboo for a team
  • Pin settings for everyone. Admins who deploy Clawboo for a team can pin settings in one file, ~/.clawboo/operator.json or the file CLAWBOO_OPERATOR_SETTINGS names, or with each key’s environment variable. Settings > System > Deployment shows a pinned setting as “Set by your administrator”. A file Clawboo cannot trust stops the server at startup, and the file never holds secrets. See Pin settings for everyone.
  • Turn the dashboard’s updates off. CLAWBOO_DISABLE_SELF_UPDATE=1 removes the update chip, stops the npm check and refuses the in-app update. Without it, a global install the server’s user cannot write, such as a root-owned /usr/local, is no longer offered the one-click update, which could only fail there.
  • Keep the browsers from starting. CLAWBOO_DISABLE_BROWSER_AUTOSTART=1, or the switch in Settings > System > Deployment, stops Playwright and Chrome DevTools starting on a new install. When the pinned version is already installed, each browser now starts from that copy instead of through npx, which kept an npm exec process running beside it and asked the registry on every start. Chrome DevTools runs with its usage statistics and update check off.
  • A release reads its own catalog. The Marketplace catalog comes from the release’s tag on GitHub instead of the main branch, so content written for a newer version no longer reaches an older install. A development build still reads main.
  • Updates keep node options. After an update from the chip, the server restarts with the node options it was started with, such as --max-old-space-size.
2026-10-21
The Board becomes a calendar and a task board at once, moving a card by hand does what it says, and a task’s cost counts every run. See Board.The Board
  • Days as rows. Status columns run across the top and days run down the side. Week is the default; Today, Month and All are a click away, or D, W, M and A. Open work sits on the day it was added, and work that reaches Done or Cancelled moves to the day it finished. Open work from before the range gathers in an Earlier row, and empty days fold into a thin line you can open.
  • Needs you lives in today. Approvals and stuck tasks sit in today’s row with Retry, Assign and Dismiss, and tasks that broke the same way share one card.
  • Routines on their days. Each routine’s upcoming runs show in To do on the days they are scheduled, from the new GET /api/schedules/occurrences. A Routines switch hides them.
  • Since you last looked, and Find. A line under the title says what changed while you were away, and changed cards carry a dot until you look. / finds any task or routine.
  • Calmer cards. One detail row: the agent, its runtime, one status mark or the cost, and when the card was added or finished. Unassigned work no longer claims to run on OpenClaw.
Moving cards by hand
  • A drop goes to the column under the pointer and nowhere else.
  • In progress takes no card by hand, because moving it there started nothing. A card left there shows in Needs you as Not started.
  • Finishing a card asks which day it finished when it sits on another day, and the board PATCH takes a completedAt. Cancelling always asks first, and cancels the plan steps that were waiting on it.
  • Moving a card an agent is working on stops that run, its late result can no longer override you, and whoever delegated the task is told.
Also in this release
  • Task costs that count every run. A task’s cost is the sum of every run, failed ones included. A run with no reported charge is priced from its token usage, or from its prompt and reply length when there is none, and marked as an estimate.
  • A quieter team chat. A teammate’s reply reaches the chat only when you addressed it or it is the lead’s summary, and Stop also stops a run that was about to start.
  • Routing by role. On the lead’s turn the roster lists each teammate’s runtime and a one-line role from its SOUL.md, so Boo Zero routes work by what a teammate does.
2026-09-28
Agents reach outside apps through connectors, team memory becomes a graph that learns, the Scheduler becomes Routines, and the Board gains a Needs you column.Connectors
  • A Connectors view in the sidebar. Curated connectors Clawboo runs itself, a longer list from the official MCP registry, and Gmail, Slack, Jira and more through Composio. Turning one on is one button, and sharing it with an agent is a line on the graph that is also the permission Clawboo enforces. See Connectors.
The graph
  • Build on the graph. Every Boo carries one visible port. Pull a thread onto another Boo to connect them, or let go on empty canvas to pick a skill, a connector or a new agent, already wired. Routes, skills and shares all come off the same way, and a share has an Undo. See Org chart and team graphs.
Memory
  • Team memory is a graph that learns. Memory is a sidebar view where facts cluster and link by similarity. Agents report whether a recalled fact helped, so a fact two teammates found useful comes first, and one that misled someone leaves automatic recall. Every recalled line carries an id the agent can cite. See Memory.
Work
  • Routines. The Scheduler is now Routines, and a routine targets a team or one agent. A team routine posts its instructions into the team chat for the lead; an agent routine files a task for that agent. See Routines.
  • Needs you, and tasks for one agent. The Board gains a Needs you column for approvals and stuck tasks, with Retry, Assign and Dismiss. New task asks for a team, then an agent, and only that agent works on it.
Platform
  • The catalog leaves the package. The agent and team catalog is downloaded instead of bundled, and grew to nineteen packs, each naming its source repository, commit and license.
  • OpenClaw 2026.9. The Gateway pin moves from ~2026.5 to ~2026.9, with the fixes that line needed.
  • Served under a path. CLAWBOO_BASE_PATH mounts the whole app under a URL prefix behind a reverse proxy.
  • In-place database upgrades. An upgrade adds new columns to an existing clawboo.db instead of failing on first use.
  • Coordination that stays awake. Task changes reach running agents through a per-agent mailbox, runs prove they are alive with regular heartbeats, and a team-chat run that ends without an answer says why.
Removed and limits
  • POST /api/onboarding/seed-native-team is removed. Deploy a team from the Marketplace, or with POST /api/teams.
  • The API has request rate limits, with a much lower one on the costliest routes, such as installing a runtime, driving the Gateway, signing in a CLI and approving a device. See Limits and defaults.
  • The Tasks MCP server caps task creation: 24 live subtasks per parent, 2 levels of nesting, and 30 top-level tasks per rolling 5 minutes.
  • A task dependency that would form a cycle is refused.
2026-07-20
A patch for running Clawboo Native teams day to day: team-chat replies stay put, status badges update live, and each runtime manages its own providers.Team chat
  • Replies stay in the transcript. Every streamed team-chat reply is saved or cleanly cleared, so a delegating agent’s message cannot vanish on the next update or on reload. A Clawboo Native one-to-one reply that fails mid-stream is kept.
  • Live Working and Idle badges. The status marks in the sidebar and in team chat update in real time for Clawboo Native and server-run team work, not only for OpenClaw.
Providers and models
  • A provider manager per runtime. Manage on Clawboo Native, OpenClaw and Hermes in the Runtimes panel lists the providers you connected, each with connect, disconnect and reconnect, plus a default-model picker for Clawboo Native. A Clawboo Native runtime with no key reads Disconnected, with a shortcut to set it up.
  • Reconnect with any provider. Reconnecting Clawboo Native works with any provider you already set up, not only Anthropic.
  • Provider first, then model. Creating a team asks for a provider (only the connected ones), then its model, and shows the exact default model that will run.
Polish
  • The Add runtimes step of setup shows your connected providers read-only, with Back to the step where you add keys.
  • The sidebar mascot’s tooltip reads “Boo Zero”, matching what clicking it opens.
2026-07-15
Native-first: the first run needs one provider key and nothing else. No OpenClaw install, no Gateway and no other CLI. OpenClaw, Claude Code, Codex and Hermes become runtimes you add to a setup that already works.Setup and providers
  • Setup starts with a key. Connect a provider, optionally add runtimes, then deploy a real team from the Marketplace. See Quickstart.
  • A Providers hub in Settings covers ten providers. Pasting a key fetches that provider’s live model list, so the model picker offers what your key can reach.
  • A first-run tour: a one-time, skippable spotlight over the sidebar. See Tour the dashboard.
Teams and chat
  • A runtime and a model for each member. Creating a team gives every member, the lead included, its own runtime and, where the runtime uses one, its own model. A choice that is not available falls back to Clawboo Native instead of blocking the deploy. See Mix runtimes on one team.
  • Clawboo Native teams chat and delegate. A Clawboo Native team sends from team chat and streams back, and its lead hands work to teammates the way an OpenClaw lead does. See Work with your team in chat.
  • Orchestration runs on the server, per team, so delegated work carries on after you close the tab.
  • Failed and abandoned delegations show on the Board instead of leaving the delegating agent waiting.
The graph
  • Every Boo carries its runtime’s mark and shows its model when expanded, and an agent’s MCP servers appear as tiles beside its skills. See Org chart and team graphs.
Resilience and security
  • A degraded Gateway no longer blocks the dashboard when you have other agents. A banner offers the recovery that applies: restart the Gateway, approve this device, or retry.
  • Hermes installs on Python 3.11 or newer instead of failing on an older system Python, and an existing codex login is recognized.
  • Binding to a non-loopback address refuses to start without an access token.
2026-06-20
Clawboo becomes an orchestrator for several agent runtimes instead of a control panel for OpenClaw. Clawboo Native agents are built in: paste a provider key and start. OpenClaw, Claude Code, Codex and Hermes join as teammates in one chat, sharing one board and one memory.
  • Five runtimes behind one adapter: Clawboo Native, OpenClaw, Claude Code, Codex and Hermes. See Choose a runtime.
  • A durable task board that agents claim work from, replacing delegation parsed out of chat text. See The board.
  • Four MCP servers, Tasks, Memory, Tools and TeamChat, that give any attached agent the board, team memory, tools and team chat. See MCP tools reference.
  • Verification: fixed checks plus an independent critic agent review some task results before they count as done. See How work is checked.
  • Guardrails: USD budgets, tool-loop breakers, limits on delegation depth and fan-out, and an audit log. See Budgets and guardrails.
  • Observability: an append-only event log, traces and an error taxonomy. See Observability.
  • Its own home in ~/.clawboo/: the database, settings, an encrypted key vault, and a folder per runtime and agent.
  • A team-task scheduler, which later became Routines.
  • A release security audit, and a loopback-only bind by default.
Removed: the delegation engine that parsed chat text, the experimental Labs UI, and the per-feature flags.Upgrading: in 0.2.0 a database schema change meant resetting the database, and a 0.1.x database could not be upgraded in place. Since 0.4.0 an upgrade adds missing columns in place. See Data and state.
2026-06-01
  • A calm welcome screen. A soft animated sky replaces the WebGL welcome animation on the setup splash and the home screen, which also makes the install smaller and spares the GPU on first load.
  • Light theme first. Fresh installs start in the light theme. Switch to dark or system afterwards.
  • Team colors. Each team picks one of 8 color palettes, and the create-team preview matches the deployed team.
  • Token usage is tracked by team and by agent.
  • On Windows, the setup’s detect step no longer hangs when a program lookup is slow.
2026-05-29
  • Light and dark themes, with a toggle that remembers your choice. See Choose a light or dark theme.
  • A design pass over surfaces, type and motion, and the radial layout for the all-teams graph (then called Atlas, now the Org chart).
  • Real provider logos and a consistent model dropdown in setup, plus fixes for setup hangs.
2026-05-20
  • Device pairing in the setup wizard. On a fresh macOS install with OpenClaw 2026.5, the approval card appears inside the wizard, with no page refresh needed.
  • Windows fixes. Launching OpenClaw and npm commands on Windows no longer fails or pops up a console window, and longer timeouts cover a slow first start.
  • Retry at the Gateway start step joins a launch already in progress instead of starting a second Gateway.
2026-05-20
  • Setup checks what is really installed on disk instead of trusting a flag saved in the browser, so you can run the install wizard again after removing OpenClaw.
2026-05-20
  • Approve this device from the dashboard. OpenClaw 2026.5 stopped pairing new devices on first connect. An Approve this device button approves it without a trip to the terminal, then retries the connection. See OpenClaw.
2026-05-20
  • OpenClaw protocol 4. Clawboo connects to OpenClaw 2026.5.18 and later (protocol 4) as well as older releases (protocol 3), and installs OpenClaw as openclaw@^2026.5.
  • Windows install support. The install and detect steps find Windows commands such as npm.cmd, so Install no longer fails with spawn npm ENOENT. The CI bundle smoke test gained a Windows leg.
2026-05-19
  • The CLI checks that a port really serves Clawboo before opening it, so it no longer opens your browser on an OpenClaw Gateway port or another app’s port. A clean-install test joins CI and the publish gate.
2026-05-19
  • Fixes Cannot GET / in 0.1.1: the bundled server serves the dashboard at /.
2026-05-19
The first real release.
  • A Marketplace with Skills, Agents and Teams tabs: 304 agents and 82 team templates.
  • The all-teams graph (then called Atlas) and team chat, with Boo Zero leading every team.
  • Delegation between agents, with multi-step plans and parallel work.
  • A setup wizard that detects, installs, configures and starts OpenClaw, then deploys a team.
  • A dashboard port that starts at 18790 and moves up as far as 18809 when a port is taken.
This release ran on OpenClaw only. 0.2.0 replaced its chat-text delegation with board-driven orchestration.
2026-02-28
  • A placeholder published to reserve the clawboo name on npm.
Last modified on November 18, 2026