MCP tools

The headless cc_server hosts a single MCP (JSON-RPC 2.0) tool registry that every client — the desktop, the web build, external MCP clients (pi, Claude Code), and the built-in agent runtime — reaches over the same server. Point any MCP client at it and it can drive the same surface the app drives itself. This reference catalogs the tools by family.

~80 first-party tools are wired into the live registry today (newsfeed, ticket reads + the typed ticket write tools create_ticket/update_ticket/assign_ticket/comment_on_ticket/close_ticket/link_tickets/…, messaging + todo_write, memory read/write + intelligence, governance, code graph, skills install/verify/pin + update, plan playbooks, agent peer messaging + delegation, the universal read, and the discovery aids search_tool_bm25 + list_my_tools). Ticket edits are typed tools — there is no CLI-in-JSON surface (ticket_cli was retired). The full registry is advertised in tools/list (no discovery gating), and external MCP servers can be bridged in.

Common parameters

`workspace_id`

Required on every tool that touches workspace-scoped data. A missing or invalid workspace_id returns an explicit error, and a value that doesn’t match the target entity’s workspace is rejected rather than silently ignored. Omit it only on the genuinely global tools (e.g. list_workspaces). When an agent is dispatched by the server, the workspace is bound to its session and forced server-side, so the agent cannot reach another workspace by passing a foreign id.

Agent tools

Tool	Description
`list_agents`	List agents in a workspace
`hire_agent`	Create a new agent (writes an `AGENTS.md` and registers it)
`update_agent`	Update an agent’s configuration
`fire_agent`	Remove an agent from the workspace
`kill_agent`	Terminate a running agent process
`get_agent_run_logs`	Get run logs for an agent

Agent collaboration tools

Tool	Description
`propose_hire`	Propose hiring a new agent (posts a hire proposal for approval)
`request_peer_review`	Ask a peer agent to review work

Skill tools

Every install and update passes through a fail-closed scan gate before any skill content reaches disk or an agent prompt.

Tool	Description
`list_skills`	List available skills
`create_skill`	Create a new skill definition
`install_skill`	Install a skill from GitHub, pinned to a commit SHA
`list_skill_updates`	List skills with a newer version available on the registry
`verify_skills`	Verify locked skills against their on-disk content hashes (drift + verdict staleness)
`pin_skill`	Record an existing on-disk skill in the lockfile

Plan tools

Tool	Description
`create_playbook`	Create a named, versioned, parameterized plan template
`run_playbook`	Run a playbook against a target, substituting its typed parameters

Agent peer messaging and delegation tools

Agents talk to each other over channels (durable, roster-visible). ask_agent is request/reply with a mandatory timeout; delegate_task is guarded by depth, cycle, budget, and autonomy limits enforced server-side.

Tool	Description
`send_to_agent`	Send a fire-and-forget message to a specific agent
`ask_agent`	Ask another agent a question and block until it replies (or times out)
`delegate_task`	Delegate a scoped task that materializes as a child ticket or plan node
`todo_read`	Read your own conversation’s task checklist
`consult_agent`	Ask another agent a question and get its answer back

Workspace tools

Tool	Description
`list_workspaces`	List all workspaces (global)
`create_workspace`	Create a new workspace (global)

Repo tools

Tool	Description
`list_repos`	List repositories linked to a workspace

Messaging tools

Tool	Description
`list_channels`	List channels in a workspace
`get_channel_messages`	Get messages from a channel
`send_channel_message`	Send a message to a channel
`todo_write`	Replace a channel’s task checklist in one call

Pull request and review tools

Tool	Description
`list_pull_requests`	List PRs for a workspace’s repos
`start_ai_review`	Start an AI-powered review run for a PR
`add_review_node`	Add a review finding (P0–P3) to a review channel
`confirm_review_node`	Confirm another reviewer’s finding
`dismiss_review_node`	Dismiss a finding (optionally recording why as a memory fact)
`dispatch_reviewers`	Dispatch matched reviewer agents into a review channel
`submit_reviewer_verdict`	Submit a reviewer agent’s ship/hold/block verdict
`finalize_review`	Finalize and roll up the review verdict
`publish_review_to_github`	Publish the review (inline comments + verdict) to GitHub

Memory tools

Tool	Description
`search_memory`	Hybrid (BM25 + vector) search over facts and policies
`propose_fact`	Propose a new memory fact in a domain
`supersede_fact`	Supersede an existing fact with a newer one
`propose_policy`	Promote facts into a normative policy
`list_policies`	List policies in a domain
`list_memory_domains`	List memory domains in a workspace
`record_observation`	Record a raw observation (later promoted to facts)
`update_my_notes`	Update the calling agent’s working-memory scratchpad
`get_my_notes`	Read the calling agent’s working-memory scratchpad
`remember`	Write to the agent’s hot working-memory tier
`consolidate_memory`	Run a consolidation (“sleep”) pass over facts
`harmonize_memory`	Reconcile a detected belief conflict
`list_memory_conflicts`	List detected memory conflicts

Governance tools

Tool	Description
`create_goal` / `list_goals` / `update_goal_progress`	Create, list, and advance per-agent/team goals
`create_approval` / `list_approvals` / `decide_approval` / `comment_approval`	Board-style approval gates and their decisions
`exit_plan_mode`	Open (and, once approved, consume) a `plan_exit` approval, flipping a channel out of plan mode
`agent_heartbeat` / `list_runtime_health` / `list_agent_presence`	Runtime liveness heartbeat, health, and presence
`get_org_chart`	Read the workspace org chart (reporting tree)
`create_work_product` / `save_work_product_revision` / `list_work_products` / `get_work_product`	Versioned agent deliverables
`create_runtime_profile` / `list_runtime_profiles`	Per-agent runtime configuration profiles

Code graph tools

Tool	Description
`search_code`	Search code symbols by query (hybrid search)
`code_symbol`	Get a symbol’s details
`code_callers`	Find callers of a symbol
`code_callees`	Find callees of a symbol
`code_impact`	Impact-radius (BFS) analysis from a symbol

Resource tools

Tool	Description
`read`	Read a file or resource via an internal URL (`pr:`, `issue:`, `gh:`, `skill:`, `rule:`, `local:`, `agent:`, `artifact:`, `memory:`)

User interaction tools

Tool	Description
`ask_user_question`	Ask the user a question (options and/or free text); blocks until answered
`request_confirmation`	Request approval for a privileged or destructive action

Discovery tools

Navigation aids over the (fully listed) catalogue. Both report which tools are callable in your current mode, so you never have to discover a restriction by calling a tool and getting rejected.

Tool	Description
`search_tool_bm25`	Find a tool by intent (BM25 over the catalogue); each hit is annotated with `callable` + a `restricted_reason` when blocked in the current mode
`list_my_tools`	The authoritative “what can I call right now” view — partitions the catalogue into `available` and `restricted` (with reasons) for your channel’s mode

Diagnostic tools

Tool	Description
`doctor`	Run agent-environment diagnostics

Ticketing tools

Tool	Description
`create_ticket`	Create a ticket
`get_ticket`	Get ticket details
`list_tickets`	List tickets in a workspace
`update_ticket`	Update a ticket’s status, title, description, priority, or labels (status accepts aliases like `in_progress`/`completed`)
`assign_ticket`	Assign a ticket to an agent
`reassign_ticket`	Reassign a ticket to a different agent
`add_ticket_collaborator`	Add a collaborator to a ticket
`comment_on_ticket`	Comment on a ticket
`link_ticket_to_pr`	Link a ticket to a pull request
`unlink_ticket_from_pr`	Remove a ticket↔PR link
`delegate_ticket`	Delegate a ticket to another agent
`fail_ticket`	Mark a ticket failed (with an error)
`close_ticket`	Close a ticket
`submit_output`	Submit the structured output payload for your current pipeline run
`link_tickets`	Create a dependency edge between two tickets (blocks / relates-to / duplicate-of)
`unlink_tickets`	Remove a dependency edge
`list_ticket_relations`	List a ticket’s dependency relations

Project tools

Tool	Description
`create_project`	Create a project
`list_projects`	List projects in a workspace
`update_project`	Update a project
`delete_project`	Delete a project
`set_ticket_project`	Move a ticket into (or out of) a project

Orchestration tools

Tool	Description
`propose_orchestration`	Emit a structured multi-agent orchestration plan for one upfront approval; validates and returns any violations

Newsfeed tools

Tool	Description
`list_feeds`	List the workspace’s RSS/Atom feeds
`list_articles`	List articles from the workspace’s feeds
`get_article`	Get a single article’s details
`set_article_read`	Mark an article read or unread
`set_article_saved`	Save or unsave an article
`refresh_feeds`	Refresh the workspace’s feeds

The calendar and meetings features are desktop-only. They are not exposed over MCP, so agents cannot read your calendar or your meeting recordings. The meeting summary that produces action items and decisions runs as a built-in meeting_summary pipeline, not as a tool agents can call directly.

Tool gating

Tools are gated by mode via ModeToolGuard:

Chat mode: full tool access.
Plan mode: read-only tools; no file writes or mutating tools.
Review mode: review-specific tools only.

A blocked call returns a message explaining the restriction and how to lift it (e.g. exit_plan_mode). Call list_my_tools at any time to see exactly which tools are available versus restricted for the current channel — gating is never a silent failure.

Destructive-action confirmation

Some tools require explicit user approval before they run. The tool returns an ApprovalPayload; the dispatcher surfaces it (via the ConfirmationPort / request_confirmation) and the client must confirm before execution proceeds. Examples include kill_agent, fire_agent, and destructive bash commands run inside the sandbox.