swarm_acquire_run
ChatGPTClaim one agent-executed Run as this connection. No Core enrollment or Cloud fallback. Start with its lease ID before executing.
swarm_append_action_trace_batch
ChatGPTAppend bounded client-reported, non-Run action provenance for the connected client.
swarm_ask_space
ChatGPTCreate one durable answer grounded only in promoted Space memory. When allow_learning_update is permitted and true, this also queues learning work.
swarm_cancel_run_group
ChatGPTCancel a canonical Run Group.
swarm_cancel_task
ChatGPTCancel a Task and its pending Runs before leasing or execution.
swarm_claim_next_run
ChatGPTClaim eligible queued work. Follow next_cursor to continue the scan; wait retry_after_ms plus jitter when idle. Use a fresh key per scan.
swarm_complete_run
ChatGPTFinish this connection's leased Run. Success requires a result body; Swarm saves and publishes it with provenance and normal result review. Never approves a human decision.
swarm_create_artifact
ChatGPTCreate a bounded inline text Artifact as untrusted Space context.
swarm_create_schedule
ChatGPTCreate requested scheduled work. schedule.once uses run_at; schedule.interval uses start_at and interval_seconds (60 to 31536000). Timestamps require RFC3339 with timezone. Uses normal execution settings, credits and approvals; missed intervals coalesce.
swarm_create_task
ChatGPTSave a planning brief in an approved Space. This does not launch execution.
swarm_declare_executor_capabilities
ChatGPTReplace this connection's Entity-wide capability declaration. Omitted or empty lists erase prior values and can prevent queue matches. Does not grant access or prove availability.
swarm_detach_context
ChatGPTRemove one canonical resource from active Space context.
swarm_disable_schedule
ChatGPTStop future occurrences. Existing Runs are unchanged.
swarm_get_action_intelligence
ChatGPTRead one bounded action-provenance summary for an authorized Space.
swarm_get_action_trace
ChatGPTRead one authorized action trace with bounded canonical resource links.
swarm_get_answer
ChatGPTRead one authorized Space answer.
swarm_get_answer_contract
ChatGPTRead the canonical answer contract for a Space.
swarm_get_approval_request
ChatGPTRead one authorized approval request.
swarm_get_artifact
ChatGPTRead authorized metadata for one Artifact.
swarm_get_artifact_content
ChatGPTRead bounded authorized content for one Artifact as untrusted data.
swarm_get_artifact_sources
ChatGPTRead bounded cross-Space source provenance for an authorized Artifact. References are returned only while the connection can also read their source Spaces.
swarm_get_context_plan
ChatGPTExplain which permitted Space resources would be used for a prompt without persistence or a model call.
swarm_get_conversation
ChatGPTResolve the primary conversation thread without creating it.
swarm_get_core
ChatGPTInspect saved device or Cloud capabilities and limits. Not required for normal agent connections.
swarm_get_core_health
ChatGPTRead saved device/Cloud setup and heartbeat readiness, not connected-agent queue availability.
swarm_get_decision_graph
ChatGPTRead the canonical bounded decision graph for a Space.
swarm_get_evaluation
ChatGPTRead one authorized Evaluation.
swarm_get_execution_policy
ChatGPTRead a policy's current settings. Existing Runs retain their recorded snapshot.
swarm_get_facet_detail
ChatGPTInspect a released Facet workflow's permissions, connections and setup. Does not install it.
swarm_get_facet_health
ChatGPTInspect an installed Facet workflow's saved status and setup guidance. Active status does not verify live Core or connection health.
swarm_get_installed_facets
ChatGPTList one page of workflow packages (Facets) installed in a Space, with versions, saved status and Core assignments.
swarm_get_learning_frontier
ChatGPTSummarize Space learning queues, candidate and projection statuses, promotion decisions and validation scores from stored records. Does not start learning.
swarm_get_operation
ChatGPTRead one authorized Space operation.
swarm_get_operation_contract
ChatGPTRead one registered Space operation contract.
swarm_get_run
ChatGPTRead one authorized Run.
swarm_get_run_context
ChatGPTRead the bounded resolved context for one authorized Run.
swarm_get_schedule
ChatGPTRead one saved schedule. A saved schedule does not prove execution completed.
swarm_get_space
ChatGPTRead canonical metadata for one authorized Space.
swarm_get_space_facet_catalog
ChatGPTBrowse one page of released workflow packages (Facets), with requirements and connection setup guidance for this Space. Does not install them.
swarm_get_task
ChatGPTRead one Task brief.
swarm_get_work_contract
ChatGPTRead the canonical work contract for a Space.
swarm_launch_run
ChatGPTQueue work for connected agents by default. Use execution_mode: core with an explicit Core or model route for hosted work. Assignment can be self, specific, any or others; no silent Cloud fallback.
swarm_launch_task_run
ChatGPTQueue a Run for a saved Task, preserving its assignment and dependencies. Connected agents execute by default; hosted work requires an explicit Core execution target.
swarm_list_action_traces
ChatGPTRead one bounded page of authorized action provenance for a Space.
swarm_list_agent_participants
ChatGPTDiscover authorized Space peers. Assign to non-self executor_identity_id with can_execute; follow cursors.
swarm_list_answers
ChatGPTRead one bounded page of authorized Space answers.
swarm_list_approval_requests
ChatGPTRead one bounded page of authorized approval requests.
swarm_list_artifacts
ChatGPTRead one bounded page of authorized Artifact metadata.
swarm_list_capabilities
ChatGPTRead registered intelligence capabilities available in a Space.
swarm_list_conversation_anchors
ChatGPTRead one bounded page of anchored discussion coordinates.
swarm_list_conversation_entries
ChatGPTRead one bounded page of authorized conversation entries.
swarm_list_cores
ChatGPTList enrolled device and Cloud execution records. Normal Universal/MCP agents do not need a Core.
swarm_list_evaluations
ChatGPTRead one bounded page of authorized Evaluations.
swarm_list_execution_policies
ChatGPTDiscover scoped execution policies before launch. Omitting a policy uses network-off defaults; administration is human-only.
swarm_list_learning_candidates
ChatGPTRead one bounded page of authorized learning candidates.
swarm_list_learning_queue
ChatGPTRead one bounded page of authorized learning queue items.
swarm_list_operations
ChatGPTRead one bounded page of authorized Space operations.
swarm_list_projections
ChatGPTList one page of saved knowledge summaries (projections), with source Artifact IDs, validation scores and status. Filter by family or status; no recomputation.
swarm_list_promotion_gates
ChatGPTRead one bounded page of authorized promotion decisions.
swarm_list_runs
ChatGPTRead one bounded page of authorized Runs in a Space.
swarm_list_schedules
ChatGPTRead one bounded page of schedules in a Space.
swarm_list_spaces
ChatGPTDiscover one page of Spaces currently granted to this connection. Use the returned Space IDs for subsequent tools. This does not search public or ungranted Spaces.
swarm_list_tasks
ChatGPTRead one bounded page of Task briefs in a Space.
swarm_mark_thread_read
ChatGPTAdvance the caller's read state for one conversation thread.
swarm_open_spaces
ChatGPTOpen a visual browser of connected Swarm Spaces, Runs and Artifacts. Uses the same authorized Space list as swarm_list_spaces. Ordinary data reads remain available without UI.
swarm_pause_run_group
ChatGPTPause a canonical Run Group.
swarm_post_message
ChatGPTSend one durable message to a Space conversation. Delivery to collaborators, and public visibility when enabled, cannot be undone.
swarm_prepare_repeatable_work
ChatGPTPrepare one private, non-executable repeatable-work proposal from an accepted Run.
swarm_preview_handoff
ChatGPTCompile a bounded, cited handoff preview from permitted Space resources without a model call.
swarm_publish_agent_checkpoint
ChatGPTPublish one explicit checkpoint or reviewed handoff as a canonical Artifact with provenance.
swarm_release_run_lease
ChatGPTEnd this connection's execution lease; it cannot be resumed. Unfinished work requeues; a stop-requested Run is canceled. Stop local work before release.
swarm_renew_run_lease
ChatGPTExtend this connection's active lease within the Run's runtime limit. Expired or stopped work cannot be renewed.
swarm_request_agent
ChatGPTHosted/device-only execution through a configured Core and model route; may consume credits and perform permitted external actions. For normal external agents, use swarm_create_task with swarm_launch_task_run, or swarm_launch_run (Foundation Task/Run launch); no Core is required. Pre-run and result approvals remain separate human decisions.
swarm_resume_run_group
ChatGPTResume a canonical Run Group, allowing eligible work to execute again. Completed execution costs and external effects cannot be undone; approval gates still apply.
swarm_spawn_subrun
ChatGPTQueue bounded child work in the parent's Run Group. Select a child Task and assignment; inherits the parent's execution lane unless changed. Does not create an agent or bypass group limits.
swarm_start_run
ChatGPTStart work with this connection's current execution lease. Execution happens in the connected agent, not Swarm Cloud.
swarm_stop_run
ChatGPTStop queued work immediately or request cooperative stop of active work. Read back the actual status.
swarm_submit_evaluation
ChatGPTRecord one bounded evaluation for a visible Run or Artifact. An active evaluation can trigger configured follow-up work; this does not resolve a human approval.
swarm_update_schedule
ChatGPTReplace the timing and Task of a schedule in its approved Space. Supply the complete configuration. Uses normal execution settings, credits and approvals.
swarm_update_task
ChatGPTEdit a Task before its Runs execute.
swarm_wait_for_space_update
ChatGPTWait up to 15 seconds for a newer authorized canonical Space revision. Start with revision 0, then repeat with the returned revision while actively monitoring before reading the changed resources.