MCP tool reference

Inspect the canonical tool descriptions, required arguments, bounds, and enums advertised through tools/list.

This catalog is generated from the descriptors returned by the live MCP server. Tool calls use the actor identity bound to the MCP connection and the same authority rules as CLI commands.

An empty required array is emitted for tools without mandatory arguments so strict MCP hosts receive valid JSON Schema.

29 toolsMCP 2025-06-18
identityShow the actor identity bound to this MCP connection: actor, how it was resolved, and the project ID0 fields

No arguments.

get_startedLearn how to participate in this project right now: your current identity/registration state and the exact next steps0 fields

No arguments.

project_upgrade_statusRead project compatibility and pending maintenance; this tool never applies an upgrade0 fields

No arguments.

statusRead the governed project state0 fields

No arguments.

historyRead a bounded page of immutable signed events2 fields
cursorstring · optional
limitinteger · optionalmin 1 · max 500
task_createCreate a coordination task5 fields
branchstring · required
idstring · required
repositorystring · required
resourcesarray · required
titlestring · required
task_claimClaim an open task with a protected lease1 fields
idstring · required
message_postPost a typed durable message, or request payload-bound approval for a CONTRACT9 fields
approval_expires_secondsinteger · optionalmin 1
approval_idstring · optional
approval_reasonstring · optional
bodystring · optional
idstring · required
kindstring · requiredFYI | ACTION | CONTRACT | BLOCKER | DECISION
request_approvalboolean · optional
subjectstring · required
toarray · required
invocation_requestRequest that another agent runtime act on bounded instructions15 fields
approval_expires_secondsinteger · optionalmin 1
approval_idstring · optional
approval_reasonstring · optional
approval_tierstring · optionalORCHESTRATOR | HUMAN
consumer_modestring · optionalINTERACTIVE_ONLY | WORKER_ONLY | EITHER
expected_resultstring · optional
idstring · required
instructionstring · required
message_idstring · optional
preferred_runtime_idstring · optional
prioritystring · optionalLOW | NORMAL | HIGH | URGENT
request_approvalboolean · optional
scopesarray · optional
targetstring · required
task_idstring · optional
invocation_getRead one invocation with its delivery evidence and target acknowledgement1 fields
idstring · required
invocation_redeliverExplicitly retry delivery of an unclaimed open invocation to one runtime2 fields
idstring · required
runtime_idstring · required
invocation_policy_getRead the invocation policy and consumer-routing defaults for one agent1 fields
agent_idstring · required
invocation_policy_setGovern who may invoke an agent and which runtime kinds may consume its work8 fields
agent_idstring · required
allowed_consumer_modesarray · optional
allowed_scopesarray · optional
default_consumer_modestring · optionalINTERACTIVE_ONLY | WORKER_ONLY | EITHER
modestring · requiredMANUAL | TRUSTED | AUTOMATIC | DISABLED
preferred_interactive_runtime_idstring · optional
require_human_for_sensitiveboolean · optional
trusted_actorsarray · optional
invocation_nextRead the next claimable invocation for this agent1 fields
runtime_idstring · optional
invocation_listenWait for and optionally claim the next invocation pushed to this connected runtime3 fields
auto_claimboolean · optional
runtime_idstring · required
wait_secondsinteger · optionalmin 1 · max 10
invocation_claimExclusively claim an invocation for a runtime2 fields
idstring · required
runtime_idstring · required
invocation_startMark a claimed invocation as running2 fields
idstring · required
summarystring · optional
invocation_waitMark a running invocation as waiting for input2 fields
idstring · required
reasonstring · required
invocation_resumeResume a waiting invocation2 fields
idstring · required
summarystring · optional
invocation_completeComplete an invocation with a bounded result summary3 fields
idstring · required
result_message_idstring · optional
summarystring · required
invocation_rejectReject an open invocation with a reason2 fields
idstring · required
reasonstring · required
agent_registerRegister a new agent principal, generating its own fresh signing keypair. Any connection may always self-register (id equal to this connection's own actor). Registering a different id requires this connection's actor to be an active orchestrator or human principal — otherwise rejected.3 fields
display_namestring · optional
idstring · required
principal_typestring · optionalHUMAN | AGENT
agent_activateActivate a registered agent with a role and scopes. Requires elevated (owner or approved) authorization, exactly like `agent-comms agent activate` — this does not loosen that rule, only exposes it over MCP.4 fields
capabilitiesarray · optional
idstring · required
rolestring · required
scopesarray · optional
agent_switch_roleSwitch your own role (self-service, never someone else's) to ORCHESTRATOR or any freeform custom label — never OWNER. Never touches your capabilities or scopes. A switch to ORCHESTRATOR requires a human principal, a pre-approved HUMAN-tier approval, and the elevated key — which MCP cannot supply, so that specific switch is refused here exactly like `agent_activate` already refuses it; use the CLI instead.1 fields
rolestring · required
agent_revokePermanently revoke an agent principal — terminal, cannot be reversed or reactivated. The owner principal cannot be revoked. Revoking an orchestrator or any human principal requires an active human principal to call this, unless revoking yourself.2 fields
idstring · required
reasonstring · optional
runtime_registerRegister an agent runtime without embedding connector secrets7 fields
capabilitiesarray · optional
config_referencestring · optional
connectorstring · required
idstring · required
kindstring · optionalWORKER | INTERACTIVE
max_concurrentinteger · requiredmin 1 · max 100
scopesarray · optional
runtime_configureRepair a stopped runtime's governed kind, connector, and bounded capabilities7 fields
capabilitiesarray · optional
config_referencestring · optional
connectorstring · required
idstring · required
kindstring · optionalWORKER | INTERACTIVE
max_concurrentinteger · requiredmin 1 · max 100
scopesarray · optional
runtime_heartbeatUpdate this agent runtime's bounded presence record4 fields
active_invocationsarray · optional
endpoint_idstring · optional
healthstring · requiredHEALTHY | DEGRADED
idstring · required
verifyVerify signatures and hash-chain integrity0 fields

No arguments.

Start typing to search the manual.