Skip to main content
Use this page when you want to add agents to your fleet: either one specialist at a time or a pre-wired team. The Marketplace is a static, committed catalog of 304 agents, 82 teams, and 30 skills you can browse offline and deploy with two clicks. The catalog itself never changes at runtime; it is codegen’d from upstream MIT-licensed repos at pinned commits (see Marketplace catalog reference). What this page covers is the dashboard surface that browses it: the three tabs, the search and filter controls, the detail modals, and the two deploy paths. The UI is MarketplacePanel; deploying funnels into the same team-create + agent-create pipeline documented in Teams.

Prerequisites

The Marketplace is browsable with no Gateway and no runtime connected; it is a static catalog rendered entirely client-side. Deploying an agent or team requires a connected runtime (Clawboo Native or OpenClaw) so the new Boo records can actually be created.

Where it lives

Open the Marketplace from the Marketplace nav button (shopping-cart icon) in the left sidebar, or press Cmd/Ctrl + 3. The panel fills the content area with a toolbar (the three tab toggles + sort), a filter bar, and a responsive card grid. The panel opens on the Teams tab by default. The toolbar shows all three tab toggles with live counts: Teams (82), Agents (304), Skills (30). Each tab keeps its own search query and filters, so switching tabs never loses your place.

The three tabs

Teams tab

The default view. Each team renders as a TeamTemplateCard showing its emoji, name, agent count, a source badge, an optional amber Synthetic pill (for the 30 codegen’d “Excellence Team” partitions), a category label, the description, and the agent roles in the roster. Buttons: Details and Deploy. The grid leads with a Curated teams banner and a Start from scratch card (deploy a blank custom team); this tab is also where the sidebar’s + create-team button lands.

Agents tab

Each agent renders as an AgentCard showing its mascot avatar, name, role, a source badge, a colored domain pill, a category label, a two-line description, and a stats line (N skills • in N teams). Two buttons: Details (opens the detail modal) and Deploy (creates the agent in a dedicated team).

Skills tab

Each skill renders as a SkillCard with a category dot, a neutral Curated tag (these are a hand-curated, first-party catalog — there is no external skill registry, fetch, or vetting behind them), a two-line description, and an Add button. When a skill is used by catalog agents, a Used by N agents link appears that cross-jumps to the Agents tab pre-searched on that skill name.
Adding a skill is different from deploying an agent. Add records a capability annotation on an existing agent in your fleet (you pick the target from a dropdown); it POSTs to /api/skills, which injection-scans the entry before recording it. The annotation surfaces on the Ghost Graph and the Capabilities dashboard — it labels intent, it does not provision a runtime tool (an agent’s executable tools come from the MCP broker). Deploying an agent or team creates new Boos.

Search and filters

Each tab has its own search box and filter pills. Search is a case-insensitive substring match; it never hits the network. The Source filter (shared by Agents and Teams) has four options: Filters compose with search: the result set is search(query) then narrowed by the active domain/category pill and the active source pill. When nothing matches, the grid shows an empty state with a Clear filters button that resets that tab’s search and pills.

The detail modals

Both Details buttons open a modal you can dismiss with Escape or the close button.

Agent detail (the zero-loss identity view)

AgentTemplateDetail shows the agent’s avatar, name, role, source/domain/category badges, the full description, a source-attribution link, clickable skill chips, “Appears in N teams” chips, and, the payoff, the agent’s full identityTemplate rendered as Markdown, labeled “Full source from <source>, preserved verbatim.” This is the zero-loss guarantee made visible: every catalog entry stores its complete original source spec (not a summary), and that exact text is what lands in the agent’s IDENTITY.md when you deploy. You read the entire spec before committing. The chips are cross-reference shortcuts:
  • Clicking a skill chip jumps to the Skills tab, pre-searched on that skill.
  • Clicking a team chip jumps to the Teams tab, pre-searched on that team.
A Deploy button at the bottom runs the same single-agent deploy as the card.

Team detail

TeamTemplateDetail shows the team’s emoji, name, source badge, category, description, source-attribution link, an optional expandable Workflow narrative, and the full roster: each agent with its avatar, role, parsed skills, and parsed @-mention routing (so you can see who delegates to whom before deploying). It also has a Deploy button.

Deploy

Both deploy paths open the same CreateTeamModal and end at the same team-create + agent-create pipeline. The difference is how many agents land and how the modal is labeled.

Deploy a single agent

Click Deploy on an AgentCard (or in the agent detail modal). Clawboo wraps that one agent into an adhoc one-agent TeamTemplate (id adhoc-<agentId>) and opens CreateTeamModal. The modal detects single-agent mode by shape (agentIds.length === 1) and relabels the Customize step; the title reads “Deploy agent” and the confirm button reads “Create agent”, and the pick step is skipped. The deploy loop is otherwise identical to a team deploy: it always creates a dedicated team for that one agent, then creates the agent inside it. (A single-agent deploy is just a one-member team.)

Deploy a team

Click Deploy on a TeamTemplateCard (or in the team detail modal). CreateTeamModal opens pre-filled with the template’s name, emoji, and color, skipping the pick step. The Customize step lets you rename/recolor; the confirm button reads “Deploy team”. On confirm it:
  1. POSTs /api/teams to create the team (with a client-minted UUID so the preview palette matches the deployed team).
  2. Calls createAgent() for each roster member, writing the verbatim identityTemplate to that Boo’s IDENTITY.md.
  3. POSTs /api/teams/:id/agents to assign each new Boo to the team.
When the modal finishes, Clawboo selects the new team and opens its group chat so you can use it immediately. See Teams for managing the team afterward (leaders, rules, color collections).

Verify it worked

  • After a deploy, the dashboard switches you into the new team’s group chat. The team appears in the team sidebar and the Ghost Graph.
  • Open the agent in the fleet; its IDENTITY.md should contain the same source text you saw in the agent detail modal.
  • After you Add a skill, a success toast confirms it (“Added … to … tool profile”), and the skill appears as a capability annotation on the agent in the Capabilities dashboard and on the Ghost Graph.

Troubleshooting

Some roster agents are missing after a deploy. A per-agent create failure does not abort the run: the loop records the name, continues through the rest of the roster, and ends with an error toast reading N of M agents created (… failed). If nothing at all was created, the modal drops back to the Customize step with “No agents could be created. Check that the runtime is connected, then retry.” and leaves the empty team row in place so you can retry or delete it. A name collision is not a failure mode: before creating anything, the deploy computes the smallest free numeric suffix and applies it to the team and every roster agent (Code Reviewer Boo 2), rewriting the @-mention routing to match, so you never need to rename an existing agent or deploy into a fresh fleet. See Teams.
A skill install returns an error toast. POST /api/skills injection-scans every entry first. A destructive / exfiltration / prompt-injection finding blocks the install with a 422 (the toast shows the finding), and the block is recorded in the governance audit log. This is the supply-chain guard; it is intentional, not a bug.
Counts won’t change after deploy. The Teams (82) / Agents (304) / Skills (30) toolbar counts are the catalog size, not your fleet. They are fixed; deploying does not grow them.

See also

Last modified on August 8, 2026