Agents

Workflow	Type	Description
agents._ft-dataset.build-examples	Linear	Build JSONL examples from completed AgentSessions
agents._ft-dataset.finalize	Linear	Mark AgentDataset as READY after successful upload
agents._ft-dataset.mark-failed	Linear	Compensation: set AgentDataset status to FAILED on build pipeline failure
agents._ft-dataset.upload	Linear	Upload built JSONL to storage and update the dataset row
agents._ft-dataset.validate	Linear	Gate and transition AgentDataset to BUILDING status
agents._infra-cost-sync.fetch-aws	Linear	Fetch AWS Cost Explorer billing data for agent sessions
agents._infra-cost-sync.fetch-do	Linear	Stub: DigitalOcean has no real-time per-resource billing API
agents._infra-cost-sync.fetch-gcp	Linear	Fetch GCP BigQuery billing-export cost for agent sessions
agents._infra-cost-sync.noop	Linear	No-op saga compensation for read-only infra-cost-sync layers
agents._infra-cost-sync.persist	Linear	Sum persisted AgentInfraCostEntry rows and update AgentSession infra_cost_usd
agents._infra-cost-sync.resolve-sessions	Linear	Resolve completed billable agent sessions for the requested date range
agents._session-launch.fetch-secrets	Linear	Validate AI provider credentials are decryptable for the session
agents._session-launch.launch-runner	Linear	Start the runner execution workflow and mark the session as RUNNING
agents._session-launch.load-mcp	Linear	Load MCP server definitions for the agent session
agents._session-launch.load-memory	Linear	Load agent memory entries and serialize them for injection into the runner
agents._session-launch.load-skills	Linear	Load agent skill definitions for the session
agents._session-launch.mark-failed	Linear	Compensation: mark the AgentSession as FAILED on runner launch failure
agents._session-launch.start-watch	Linear	Start the agent session watch workflow to track runner lifecycle
agents._session-launch.validate-create	Linear	Validate agent config and create AgentSession in PENDING status
agents.agent-config.clone	Linear	Clone an existing agent configuration
agents.agent-config.create	Linear	Create a new agent configuration
agents.agent-config.delete	Linear	Soft-delete an agent configuration
agents.agent-config.draft-from-description	Atomic	Draft a complete agent config (name, system prompt, model, memory) from a human-typed job description. Wraps ai.chat.complete with an internal system prompt that never leaves the server.
agents.agent-config.guidance-attach	Linear	Pin an exact trusted guidance-package version to an agent config.
agents.agent-config.guidance-detach	Linear	Detach an exact guidance-package version from an agent config.
agents.agent-config.mcp-attach	Linear	Attach an MCP server to an agent configuration
agents.agent-config.mcp-detach	Linear	Detach an MCP server from an agent configuration
agents.agent-config.purge	Linear	Permanently delete a soft-deleted agent configuration with no sessions
agents.agent-config.skill-attach	Linear	Attach a skill to an agent configuration
agents.agent-config.skill-detach	Linear	Detach a skill from an agent configuration
agents.agent-config.update	Linear	Update fields on an existing agent configuration
agents.budget-check	Linear	Check cumulative session cost against the budget and enforce limits
agents.cost-entry-record	Linear	Append a token-cost ledger row for a single LLM step within an agent session
agents.end-user.ask	Linear	Ask the app's agent a question. Runs a real agent session under the calling end-user's own authority: its tools are exactly the compositions this app exposed to end-users, every tool call is re-authorized, and the answer returns in the terminal state.
agents.end-user.ask-stream	Linear	Ask the app's agent a question and stream the answer token-by-token. Same bounded end-user session as agents.end-user.ask (its tools are exactly the app's exposed compositions, every call re-authorized); the answer also returns in the terminal state.
agents.ft-dataset.build	Dag	Build and upload a fine-tuning dataset from completed agent sessions
agents.ft-dataset.create	Linear	Create a new fine-tuning dataset export definition for the organization
agents.ft-dataset.delete	Linear	Soft-delete a fine-tuning dataset record
agents.ft-dataset.validate-format	Linear	Sample completed sessions and validate they render correctly in the dataset's export format
agents.infra-cost-sync	Dag	Fetch and persist infrastructure costs for billable agent sessions
agents.mcp-server.call-tool	Linear	Execute an MCP tool through a registered server
agents.mcp-server.discover-auth	Linear	Discover an MCP server's OAuth authorization server (RFC 9728/8414) and stage a connection for consent. Reuses an existing DCR client and user grant; pass reauthorize=true only to rotate the client (wipes tokens).
agents.mcp-server.refresh	Linear	Refresh the tool catalog for an existing MCP server
agents.mcp-server.register	Linear	Register a new MCP server
agents.mcp-server.test-connection	Linear	Test reachability and credentials for a registered MCP server
agents.mcp-server.update	Linear	Update a registered MCP server (url, description, transport, credential link)
agents.memory-distill	Linear	Distill a finished agent session into memory entries, gated on the config's memory_write_enabled flag
agents.memory-load	Linear	Load agent memory entries for a given agent config and task
agents.memory-prune	Linear	Prune excess memory entries for an agent config, removing lowest-importance oldest first
agents.memory-save	Linear	Save new memory entries for an agent config
agents.metrics-event-record	Linear	Append a typed event to an agent session event log
agents.metrics-memory-record	Linear	Append a memory query record to an agent session
agents.metrics-step-close	Linear	Close a step metric row with token counts and latency
agents.metrics-step-open	Linear	Open a step metric row for a single LLM call within an agent session
agents.metrics-tool-record	Linear	Append a tool call record to a session and emit a Lumen tool span
agents.model-capacity.acquire	Linear	Atomically reserve one agent model call against admitted RPM and TPM
agents.org-settings.upsert	Linear	Create or update agent org settings for an organisation
agents.price-catalog-sync	Linear	Sync the built-in model pricing catalog to the global AgentPriceConfig table
agents.price-config.upsert	Linear	Create or replace an organisation-specific model price override
agents.runner-group.disable	Linear	Disable agent support on a runner group, guarded by an active-session check
agents.runner-group.enable	Linear	Enable agent support on a runner group that is in ACTIVE status
agents.session-cancel-record	Linear	Set agent session status to CANCELLED
agents.session-create	Linear	Create a new agent session and open a Lumen observability trace
agents.session-fetch-attachments	Atomic	Read-only: resolve StorageObject attachments for an agent session into a metadata manifest; presigned URLs are redacted from workflow state
agents.session-finalize	Linear	Aggregate metrics and mark an agent session as COMPLETED
agents.session-finalize-cancel	Linear	Mark an agent session as CANCELLED and close the Lumen trace
agents.session-finalize-failure	Linear	Aggregate metrics and mark an agent session as FAILED
agents.session-heartbeat	Linear	Update the last heartbeat timestamp on an agent session
agents.session-heartbeat-rate	Linear	Compute heartbeat age and emit a health metric for a running agent session
agents.session-infra-cost-fetch	Linear	Return the summed infrastructure cost for an agent session from AgentInfraCostEntry rows
agents.session-infra-cost-reconcile	Dag	Re-observe and persist infrastructure cost for one terminal agent session using server-side runner-group billing configuration
agents.session-launch	Dag	Launch a new agent session: validate, create, load context, start runner, start watch
agents.session-notify	Linear	Send a Slack notification for an agent session lifecycle event
agents.session-rate	Linear	Record a quality rating for a completed agent session
agents.session-restore	Linear	Build a restore context payload from an existing agent session
agents.session-stop	Linear	Signal a running agent session to cancel (stop request)
agents.session-sweep-orphaned	Linear	Fail non-terminal agent sessions whose runner stopped reporting and whose per-session watch is no longer polling
agents.session-watch	Linear	Poll runner workflow status and finalize the agent session when done
agents.skill-package.import	Linear	Import a provider-blind guidance package pinned to an exact Git object.
agents.skill-package.revoke	Linear	Revoke a guidance-package version for new sessions without deleting history.
agents.skill-package.sync	Linear	Sync an exact new guidance-package revision without mutating old versions.
agents.skill-package.trust	Linear	Trust a validated, artifact-backed guidance package.
agents.skill.create	Linear	Create a new agent skill
agents.skill.delete	Linear	Soft-delete an agent skill
agents.skill.update	Linear	Update fields on an existing agent skill
agents.skill.validate	Linear	Validate an agent skill's backend connectivity and update last_validated_at
agents.software-delivery.cleanup-minimize	Linear	Minimize Agent-owned prompt or tool metadata under one exact leased software-delivery cleanup task
agents.tool-call.execute	Linear	Execute one model-requested tool call through the canonical workflow boundary
agents.tool-catalog.build	Linear	Build the exact OpenAI-compatible tool catalog an agent session may expose to the LLM
agents.tool-policy.evaluate	Linear	Evaluate whether a model-requested tool call may run or requires approval
agents.webhook-trigger	Linear	Evaluate and dispatch outbound webhooks for an agent session event