MCP tool index

Generated

This page is generated from the MCP server source. Do not edit it by hand — run node scripts/gen-docs.mjs. For the full parameter reference of each tool, see MCP server reference.

The Agentweaver MCP server exposes 131 tools across 15 categories. This index is the authoritative list of tool names and one-line descriptions, derived directly from the [McpServerTool] attributes in the server source.

MCP tool implementations URI-escape every route path parameter before calling the Agentweaver API. Segments such as project_id, run_id, agent_name, and task or workflow ids are encoded with Uri.EscapeDataString() so crafted ids cannot inject ../ or otherwise change the API path. Query-string parameters keep their normal query encoding.

Error responses

All MCP tool failures surface a structured JSON message:

{
  "error": "Project 'demo' not found.",
  "hint": "Call project_list to see available projects."
}

Common mappings include Agentweaver sign-in guidance for 401s, resource-specific list/read hints for 404s, review-state guidance for 409s, and diagnostics_get retry guidance for timeouts.

Addressed Message

Tool	Description
agent_message_acknowledge	Acknowledge a delivered message as received; does not change task or decision state.
agent_message_claim	At a recipient turn boundary, lease the oldest pending addressed message.
agent_message_deliver	Record delivery only after presenting the claimed message at a safe turn boundary.
agent_message_get	Get one addressed message's state and correlation.
agent_message_list	List addressed messages and delivery diagnostics visible to this caller.
agent_message_send	Persist an addressed message to one teammate's active run; acknowledgment only confirms receipt.

Backlog

Tool	Description
backlog_archive_task	Archive a backlog task off the active board. Claimed tasks also archive their linked coordinator run card.
backlog_capture_task	Capture a new task into the project backlog.
backlog_decompose_spec	Decompose a workspace spec file into proposed backlog tasks for a project. Reads a markdown file from the project's workspace, runs AI decomposition, and returns proposed items for review. Use confirm=true to create the tasks, confirm=false for preview only. Results are capped at 50 items.
backlog_delete_task	Delete a backlog task. Fails with 409 if the task has already been claimed.
backlog_edit_dependencies	Atomically add, remove, or replace a task's prerequisites at an expected project graph revision; preview=true does not mutate.
backlog_edit_task	Edit the title and/or description of a backlog task.
backlog_get_board	Get the full Kanban board for a project: Backlog, Ready, Problems, Human Review, Active, and Done.
backlog_get_dependency_revision	Get the project dependency graph revision for a safe edit or preview.
backlog_get_settings	Get the per-project backlog pickup settings (max_ready_per_heartbeat, pickup_autopilot, pickup_auto_approve_tools).
backlog_get_task	Get one enriched backlog task, including blocking dependency status.
backlog_get_workflow_stages	Get the ordered canonical run-bucket definitions for a project (Problems, Human Review, Active, Done).
backlog_move_to_backlog	Move a task from Ready back to Backlog, optionally at a specific position.
backlog_move_to_ready	Move a task from Backlog to Ready, optionally at a specific position.
backlog_reorder_task	Reorder a task within its current bucket (Backlog or Ready) to a new zero-based position.
backlog_set_settings	Set the per-project backlog pickup settings. max_ready_per_heartbeat must be between 1 and 20.
send_all_backlog_to_ready	Bulk-promote all Backlog tasks to Ready in one atomic operation. Appends them after any existing Ready tasks, preserving relative order. Idempotent — safe to call on an empty backlog.

Blueprint

Tool	Description
blueprint_generate	Start durable asynchronous Blueprint generation. Returns a job id and status/result/cancel/retry URLs immediately; use the Blueprint generation job tools to follow it.
blueprint_generation_cancel	Cancel an authorized queued or running Blueprint generation job.
blueprint_generation_result	Get the immutable Blueprint artifact for a completed generation job.
blueprint_generation_retry	Retry an authorized failed or cancelled Blueprint generation job without creating another artifact identity.
blueprint_generation_status	Get the authorized status of a durable Blueprint generation job.
list_blueprints	List the predefined Agentweaver blueprints. Each blueprint specifies a team roster, workflow, review policy, and sandbox profile ready to apply at project creation.
validate_blueprint	Validate a blueprint object against the schema and role constraints. Returns valid:true with an empty errors array on success, or valid:false with a list of validation errors.

Catalog

Tool	Description
catalog_list_roles	List all available agent roles from the catalog.
catalog_list_scenarios	List all available casting scenario templates.

Coordinator

Tool	Description
coordinator_children_get	List child runs from the same plan selected by coordinator_work_plan_get: an exact coordinator run plan or the latest persisted child-work node plan for a parent workflow run. Includes parent workflow correlation, branch node ids, persisted ordinals, subtask, agent, model, and child-run status. Empty when nothing has been dispatched.
coordinator_outcome_spec_confirm	Confirm the drafted outcome spec for a Coordinator run, resuming the suspended run past the confirmation gate.
coordinator_outcome_spec_get	Get the current persisted outcome spec for a Coordinator run.
coordinator_outcome_spec_revise	Request a revision of the drafted outcome spec for a Coordinator run. The coordinator re-drafts using the feedback and re-suspends at the confirmation gate.
coordinator_start	Start a Coordinator orchestration for a project from a plain-language goal. Optional per-run approval policy can auto-approve repository-defined safe tools and enable autopilot; destructive, privileged, preview, secret, and other network approvals remain gated.
coordinator_steer	Steer a Coordinator run. Use 'stop' to cancel active subagents immediately; 'redirect' or 'amend' to inject guidance at the targeted subagent's next turn boundary; or a recovery verb (e.g. 'recover') to reset blocked/failed/parked subtasks and auto-resume the dispatch loop. Omit target_child_run_id to broadcast to every active child. instruction is required for redirect/amend and optional for stop/recovery verbs. Pause is not supported.
coordinator_work_plan_get	Get a Coordinator work plan, including workflow parent/resume correlation, ordered joined output, branch node ids, persisted ordinals, status, child run ids, and dependency edges. A parent workflow run selects its latest persisted child-work node plan; an exact coordinator run id selects its own plan. Returns null when no work plan has been drafted yet.
orchestration_topology	Combine two authenticated work-plan and children reads for an exact coordinator run or a parent workflow run's latest persisted child-work node plan. The reads are not an atomic snapshot. For the live graph, point run_watch at the coordinator run id and consume its coordinator.topology, subtask.*, and coordinator.steering events.

Diagnostics

Tool	Description
diagnostics_get	Get a real-time system diagnostics snapshot: API version, process uptime, project/run counts, heartbeat state, and checkpoint GC state.
heartbeat_status	Get the current coordinator heartbeat service status: enabled flag, interval, last tick time, and service state (running / waiting_first_tick / disabled).
run_failure_diagnostic	Explain a failed run with bounded observed facts, attributable interpretations, unknowns, evidence completeness, the effective denial gate when recorded, and safe non-mutating next actions.

GitHub Auth

Tool	Description
github_repo_app_authorization_status	Poll the current human's Repo App browser authorization transaction. Returns only pending, completed, failed, or expired.
github_repo_app_connect	Begin the current human's Repo App authorization. Returns an opaque transaction ID, a browser URL, and expiry. Open browser_url; if Agentweaver asks for Entra sign-in, finish it in that browser to resume this handoff. After GitHub completes, return to MCP and poll github_repo_app_authorization_status. No credential, OAuth state, callback cookie, or final callback data is returned.
github_repo_app_disconnect	Disconnect the current human's Repo App authorization. This de-privileges the current human and invalidates outstanding authorization transactions.
project_copilot_app_authorization_status	Poll the initiating human's project-bound Copilot App browser authorization. Returns only pending, completed, failed, or expired.
project_copilot_app_connect	Begin an Owner-authorized, project-bound Copilot App connection. Returns an opaque transaction ID, browser URL, and expiry. Open browser_url; if Agentweaver asks for Entra sign-in, finish it in that browser to resume this handoff. After GitHub completes, return to MCP and poll project_copilot_app_authorization_status. No credential, OAuth state, callback cookie, repository, installation, permission, or final callback data is returned.
project_copilot_app_disconnect	Disconnect a project Copilot App binding. The backend allows this de-privileging operation only to an authorized human project Owner or platform administrator.
project_github_capability_status	Get redacted interactive, unattended, and repository readiness for a project. Status distinguishes unattended_ready, interactive_ready, repository_ready, reauthorization_required, and unavailable. No GitHub identities, credentials, installations, repositories, or permissions are returned.

Memory

Tool	Description
decision_compare	Retrieve two immutable decision revisions for comparison.
decision_create	Create a team decision directly (coordinator path).
decision_history	List immutable revisions for a decision.
decision_inbox_list	List inbox entries for a project.
decision_inbox_merge	Merge a pending inbox entry into team decisions.
decision_inbox_reject	Reject a pending inbox entry.
decision_inbox_submit	Submit a decision or learning to the agent inbox.
decision_list	List team decisions for a project.
decision_restore	Restore a prior decision snapshot as a new pending revision.
decision_update	Update a decision's status, content, or rationale.
memory_compare	Retrieve two immutable memory revisions for comparison.
memory_export	Export project memory to .squad/ and .agentweaver/context/ files and report the paths written.
memory_get	Get a single memory entry.
memory_history	List immutable revisions for a memory entry.
memory_import	Import .squad/decisions/inbox/*.md files into the project memory DB.
memory_list	List memory entries for a specific agent.
memory_record	Add a memory entry for an agent.
memory_restore	Restore a prior memory snapshot as a new pending revision.
memory_search	Cross-agent memory search across the whole project.
memory_update	Update memory with optimistic concurrency; approved content becomes pending.
session_current	Get the current open session for a project.
session_start	Start a new work session for a project.
session_update	Update the current session's focus, summary, or end it.
squad_decide	Submit a team decision to the decision inbox from a squad agent.

Project

Tool	Description
github_repository_selection_issue	Mint a short-lived, single-use repository selection code for one full_name returned by github_repository_selections_list. Pass only the returned code to project_create; never pass a repository URL or identifier.
github_repository_selections_list	List the signed-in caller's authorized GitHub repositories as bounded, redacted metadata. Choose one full_name from this result, then call github_repository_selection_issue before project_create with origin 'github'.
project_configure	Configure the AI model provider and generation-model settings for a project.
project_create	Create a new Agentweaver project. When origin is 'github', repository_selection_code is required; first use github_repository_selections_list and github_repository_selection_issue with the same caller. Supply blueprint_id to apply a predefined blueprint, or supply blueprint to apply an inline blueprint; the two options are mutually exclusive.
project_delete	Delete a project by ID.
project_get	Get a project by ID.
project_list	List all Agentweaver projects.
project_list_runs	List all runs for a project.
project_rename	Rename an existing project.

Run

Tool	Description
run_archive	Archive a run off active project board/list projections.
run_execution_identity	Get the authorization-filtered execution identity, delegation, retry, backend, permission-binding, and tool-decision record for a run.
run_get_file	Get the content or diff of a specific file changed by a run.
run_output_compare	Compare exact retained file identities between two output revisions of one run.
run_output_file	Read retained exact file bytes (base64), independent of the current branch or worktree.
run_output_history	List immutable output revisions for a run, including collective output identities.
run_output_revision	Read the exact immutable output revision and retained file inventory.
run_retry	Retry a failed run, or only an exact unlaunched in-progress composed recovery. Eligible coordinators, pre-dispatch composed failures, and that narrow in-progress recovery resume the same run ID; other failed-run retries create a fresh run. Arbitrary active runs cannot be retried.
run_review	Approve or reject a run that is awaiting review.
run_show_artifacts	List the files changed by a run.
run_status	Get the current status of a run.
run_submit	Legacy compatibility alias that starts a coordinator run directly in direct mode. Prefer run_task for the common one-call flow, or coordinator_start for full manual control.
run_task	Start a coordinator run once, poll the same run until completion, a proven human review or confirmation gate, or timeout. For automated child waits and timeouts, continue with run_status or run_watch; never rerun run_task to resume.
run_watch	Watch a run live, streaming progress until completion.
start_preview	Register a live browser preview for a web server the agent has ALREADY started and verified inside a run's sandbox pod. Call this AFTER your server is running and responding (e.g. you confirmed curl http://localhost:PORT/ succeeds) — pass the exact port it listens on (e.g. 3000). If observe_bound_port returned a session_id, pass it so the server can verify the process is still healthy. You MUST call this whenever you start any server so the user gets a live preview link. Routes through a human-in-the-loop approval gate; returns the public HTTPS preview_url once approved. Do not finish the task without registering the preview for any server you started.

Sandbox Policy

Tool	Description
sandbox_policy_get	Get a configured repository sandbox policy or a run's effective permission inspection.
sandbox_policy_set	Set the sandbox policy for a repository.

Skill

Tool	Description
skill_assign	Assign a catalog skill to an agent. At prompt-assembly time only assigned skills appear for that agent (progressive disclosure).
skill_assignments_list	List all skill→agent assignments in a project.
skill_create	Create or update a manual standards-compatible SKILL.md catalog skill. The name must be a lowercase kebab-case command slug.
skill_defaults_apply	Apply a matching skill_defaults_preview atomically. A stale digest is rejected; preview again before retrying.
skill_defaults_preview	Preview explicit bundled role-skill defaults for a confirmed project team. Returns a digest required by skill_defaults_apply; makes no changes.
skill_delete	Delete a catalog skill and all of its agent assignments.
skill_generate	Generate an unsaved SKILL.md draft server-side from a natural language description. Review the draft, then call skill_create to persist it.
skill_get	Get a single catalog skill including SKILL.md instructions and bundled resources.
skill_import	Import selected skills from owner/repo, https://github.com repo/tree/blob URLs, or raw https://raw.githubusercontent.com SKILL.md URLs. Idempotent by content hash. Locations are REQUIRED when a source contains multiple skills; omitting locations works only when the source has a single skill.
skill_import_preview	Preview candidate skills from owner/repo, https://github.com repo/tree/blob URLs, or raw https://raw.githubusercontent.com SKILL.md URLs, without importing.
skill_list	List catalog skills for a project with their agent assignments and status.
skill_marketplace_browse	Browse or search a curated marketplace (paginated) without changing the project catalog.
skill_marketplace_import	Import selected candidates from a curated marketplace through the normal repository-import pipeline.
skill_marketplace_source_add	Add a project-scoped skill marketplace source by public GitHub repo URL or owner/repo. Subpath is optional (auto-detected when omitted). The source is then browsable/importable by its name.
skill_marketplace_source_remove	Remove a project-scoped marketplace source by name. Administrator-curated config marketplaces cannot be removed.
skill_marketplace_sources_list	List all skill marketplaces available to a project: administrator-curated config marketplaces plus the project's own URL-added sources.
skill_marketplaces_list	List enabled administrator-curated skill marketplaces.
skill_sync	Discover and sync skills already present in the project's connected repository (.github/skills, .copilot/skills, .claude/skills, .agents/skills). Idempotent; marks vanished skills as missing.
skill_unassign	Remove a skill assignment from an agent.

Team

Tool	Description
team_cast	Cast the team for a project. Can create a proposal, confirm an existing proposal, or create+confirm in one step. If confirm_proposal_id is provided, confirms that proposal. Otherwise creates a new proposal with the given goal. Set confirm=true to automatically confirm the new proposal.
team_get	Get the current team composition for a project.
team_member_add	Add a new member to the project team.
team_member_get_charter	Get the charter document for a team member.
team_member_retire	Remove (retire) a member from the project team.

Workflow

Tool	Description
workflow_generate	Generate a new workflow definition from a natural language description, including schedule or event triggers when the description asks for them. Returns YAML draft — not yet saved. Use workflow_save to persist. Publication requests fail with an unsupported_capability response; they are never converted to agent prompts. Generated roles bind to confirmed team members; missing roles return workflow_team_binding_required with unresolved_roles. The agent can inspect the YAML before saving.
workflow_get	Get the full definition of a single workflow by ID, including its nodes, edges, and trigger. MCP has no separate trigger-configure tool yet.
workflow_save	Save a workflow YAML to the project workspace. This is the current MCP write path for trigger changes; there is no separate workflow_set_trigger tool. Validates and dry-run binds before saving. Returns the parsed workflow definition.
workflows_list	List all discovered workflow definitions for a project, including their validation status, effective default, and any configured trigger.
workflows_sync	Re-read the project's workflow definitions from disk, refreshing the in-memory registry. Returns the updated workflow list.

Workspace

Tool	Description
get_project_workspace_file	Get the content of a file in a project workspace at a given ref. Defaults to the base branch when ref is omitted.
list_project_workspace	List the flat file tree for a project workspace at a given ref. Defaults to the base branch when ref is omitted.
list_project_workspace_refs	List browsable project git refs with their resolved commit revisions: the base branch and any active run worktrees.