Tools: workflows

41 tool(s) in this category.

agent_blueprint.get

Get Agent Blueprint

Fetch one agent_blueprint by id, including the full DSL content_json. The id is either a starter slug, a starter's deterministic UUID, or a tenant blueprint UUID; the reader checks code-defined starters first, then falls through to a DB lookup. Use when inspecting steps + declared inputs of a specific recipe. Recipes for reference; walk steps yourself using primitives. Direct execution from MCP is not exposed; see SERVER_INSTRUCTIONS for the rule. Returns null when nothing matches.

• HTTP: GET /agent-blueprints/:id
• Required scopes: artifacts:read
• Required role: viewer
• Stability: stable
• Version: 1.0

Input

Field	Type	Required	Description
id	string	yes

agent_blueprint.list

List Agent Blueprints

List every agent_blueprint accessible to this business: Amdahl-shipped starters (forkable recipes in code) first, then the tenant's own DB-stored blueprints. Use when discovering which workflow recipes are available to read, fork, or compose. Recipes for reference; walk steps yourself using primitives. Direct execution from MCP is not exposed; see SERVER_INSTRUCTIONS for the rule. Each entry carries is_starter + authored_by. Fetch the full DSL via agent_blueprint://<id>.

• HTTP: GET /agent-blueprints
• Required scopes: artifacts:read
• Required role: viewer
• Stability: stable
• Version: 1.0

Input

Field	Type	Required	Description
status	string	no
limit	integer	no
offset	integer	no

agent_profile.list

List Agent Profiles

Registered agent profiles - the specialist profiles an agent can delegate to (researcher, content_writer, orchestrator). Use when deciding whether to delegate a task and to which profile, or when planning a multi-step content or research task. Returns id, name, description, producesArtifactType, and inputParamsSchema per profile; the schema tells the caller exactly which input_params keys the profile consumes.

• HTTP: GET /agent-profiles
• Required scopes: workflows:read
• Required role: viewer
• Stability: stable
• Version: 1.0

Input

No structured input.

agent_run.get

Get Agent Run

Fetch a single agent run by UUID, including step_data events, the full conversation log, token usage, and final status. Use when debugging an individual turn, rendering tool traces, or auditing an agent trace. Returns null when the run does not exist or belongs to a different business.

• HTTP: GET /agent-runs/:id
• Required scopes: workflows:read
• Required role: viewer
• Stability: stable
• Version: 1.0

Input

Field	Type	Required	Description
id	string	yes

agent_run.list

List Agent Runs

List agent runs (individual turn executions) for this business, newest first. Use when rendering per-turn activity, filtering by run status (queued|running|complete|error|canceled), a specific profile_id, or a user_id. Returns lean summaries; call agent_run://<id> for the full row.

• HTTP: GET /agent-runs
• Required scopes: workflows:read
• Required role: viewer
• Stability: stable
• Version: 1.0

Input

Field	Type	Required	Description
status	string	no
profile_id	string	no
user_id	string	no
limit	integer	no
offset	integer	no

agents.backtest_blueprint

Backtest Blueprint Over a Range

Launch a range-sweep backtest of a saved agent_blueprint: fire a series of as-of runs across a date range at a cadence (monthly|weekly), grouped into one sweep. Use when an operator wants to see how a blueprint would have performed over time, optionally cumulative (each run seeds off the prior). Every run is forced read-only. Returns the backtest_id; poll blueprint_backtest://<id> for progress.

• HTTP: POST /agent-blueprints/:id/backtest
• Required scopes: workflows:delegate
• Required role: editor
• Stability: stable
• Version: 1.0
• Tags: blueprints, write, workflow

Input

Field	Type	Required	Description
id	string	yes	Starter slug / UUID or tenant agent_blueprint artifact id.
start	string	yes	ISO-8601 date for the oldest end of the range (inclusive).
end	string	yes	ISO-8601 date for the newest end of the range (inclusive).
cadence	`enum("monthly"	"weekly")`	yes
cumulative	boolean	no	When true, runs the sweep sequentially so each run can seed off the prior run output. Default false (independent, parallel).
inputs	object	no	Inputs forwarded to every child run, keyed by the blueprint's declared input names.
acknowledge_large	boolean	no	Confirms a sweep whose run count is above the soft threshold. Required when preview reports requires_acknowledge.

agents.cancel

Cancel Agent Session

Cancel an in-flight agent run. Use when the user wants to stop a running agent before it finishes, or abandon a run that has gone sideways. Cascades to any sub-agents spawned from this run via agents.start — every live descendant is cancelled atomically. The runner bails at the next iteration boundary after observing the status change. Already-completed runs cannot be canceled. Pass a reason to record WHY the run was aborted in the audit trail.

• HTTP: POST /agents/:session_id/cancel
• Required scopes: workflows:write
• Required role: editor
• Stability: stable
• Version: 1.0
• Tags: destructive

Input

Field	Type	Required	Description
session_id	string	yes
reason	string	no	Optional operator note describing why the run was canceled.

agents.fork_blueprint

Fork Agent Blueprint

Duplicate a starter or tenant blueprint into the caller's namespace as a new draft artifact. Use when forking an Amdahl-shipped starter to customize it, or duplicating an existing tenant blueprint as a versioned variant. Source is either a starter slug (e.g. plan-and-draft-window), a starter's deterministic UUID, or a tenant blueprint UUID. Returns the new artifact id. Forks start at version 1.0.0, status draft, visibility private; promote via artifacts.update when ready.

• HTTP: POST /agent-blueprints/fork
• Required scopes: workflows:delegate
• Required role: editor
• Stability: stable
• Version: 1.0
• Tags: blueprints, write, fork

Input

Field	Type	Required	Description
source	string	yes	Slug or UUID of the source blueprint. Resolves against starters first, then tenant DB.
new_slug	string	no	Optional slug for the fork (3-60 chars, /^[a-z0-9-]+$/). Auto-generated as <source>-fork when omitted.
new_name	string	no	Optional display name for the fork. Defaults to Copy of <source name>.

agents.preview_backtest

Preview Backtest Sweep

Dry-run a range-sweep backtest WITHOUT firing anything: given a blueprint, a date range, and a cadence (monthly|weekly), return the ordered as-of dates the sweep would run at plus the count and cap signals. Use when rendering the Run modal cost-preview so the operator confirms N before launching. Nothing is written and no runs are created.

• HTTP: POST /agent-blueprints/:id/backtest/preview
• Required scopes: workflows:delegate
• Required role: editor
• Stability: stable
• Version: 1.0
• Tags: blueprints, workflow

Input

Field	Type	Required	Description
id	string	yes	Starter slug / UUID or tenant agent_blueprint artifact id.
start	string	yes	ISO-8601 date for the oldest end of the range (inclusive).
end	string	yes	ISO-8601 date for the newest end of the range (inclusive).
cadence	`enum("monthly"	"weekly")`	yes

agents.resume

Resume Agent Session

Resume an agent run that is paused on pending user input. Use when replying to an outline approval prompt, a clarification question, or a continue/finish decision. The input payload must match the run's pending_input_schema. After a successful resume the run transitions back to running and proceeds from where it paused.

• HTTP: POST /agents/:session_id/resume
• Required scopes: workflows:write
• Required role: editor
• Stability: stable
• Version: 1.0
• Tags: destructive

Input

Field	Type	Required	Description
session_id	string	yes
input	object	yes	The user's resume payload. Must match the run's persisted pending_input_schema.

agents.run_blueprint

Run Agent Blueprint

Trigger a headless run of a saved agent_blueprint on the Claude Agent SDK. Use when launching a blueprint by id with explicit inputs (optionally as a backtest via as_of). The platform runs it out-of-process: the operation validates the blueprint, opens a blueprint_runs audit row, fires the runner asynchronously, and returns the blueprint_run_id plus a preview of the rendered prompt. Poll blueprint_run://<id> (GET /blueprint-runs/:id) or the SSE stream for progress + outputs.

• HTTP: POST /agent-blueprints/:id/run
• Required scopes: workflows:delegate
• Required role: editor
• Stability: stable
• Version: 1.0
• Tags: blueprints, write, workflow

Input

Field	Type	Required	Description
id	string	yes	Starter slug / UUID or tenant agent_blueprint artifact id to run.
inputs	object	no	Resolved input values keyed by the blueprint's declared input names. Required inputs without defaults must be supplied.
as_of	string	no	Optional ISO-8601 date for a backtest. When set, the run is forced read-only and reads tenant data as of that date.

agents.start

Start Agent Session

Start a new agent run. Use when delegating a bounded, heavy task to a specialist profile (content_writer, researcher) or kicking off a one-shot copilot run. Returns a session id (run id) plus the underlying ephemeral conversation id. Poll agents.status or subscribe to SSE to monitor progress. Set async=false to block up to 60s. For multi-step content composition, read the matching blueprint recipe and perform its steps. For multi-turn chat, prefer conversations.* tools.

• HTTP: POST /agents/run
• Required scopes: workflows:delegate
• Required role: editor
• Stability: stable
• Version: 1.1
• Tags: agents, write, delegation

Input

Field	Type	Required	Description
profile_id	string	yes	Registered agent profile id (e.g. content_writer).
task	string	yes	Natural-language task description used as the initial user message.
input_params	object	no	Structured task parameters consumed by the profile (channel, content_type, project_id, etc.).
async	boolean	no	If true (default), enqueue and return immediately. If false, wait up to 60s for completion.

blueprint_backtest.get

Get Backtest Sweep

Fetch a single range-sweep backtest by id, with its as-of points resolved against their child blueprint runs (per-point run status) and a rolled-up status summary. Use when rendering a sweep progress timeline or auditing a finished backtest. Returns null when the sweep does not exist or belongs to a different business.

• HTTP: GET /blueprint-backtests/:id
• Required scopes: workflows:read
• Required role: viewer
• Stability: stable
• Version: 1.0

Input

Field	Type	Required	Description
id	string	yes

blueprint_backtest.list

List Backtest Sweeps

List range-sweep backtests (each a series of as-of blueprint runs across a date range) for this business, newest first. Use when rendering the backtest-history list, filtering by blueprint_id, or auditing past sweeps by status. Returns lean sweep rows; fetch blueprint_backtest://<id> for the per-point run rollup.

• HTTP: GET /blueprint-backtests
• Required scopes: workflows:read
• Required role: viewer
• Stability: stable
• Version: 1.0

Input

Field	Type	Required	Description
blueprint_id	string	no
status	string	no
limit	integer	no
offset	integer	no

blueprint_output.diff

Diff Blueprint Output Versions

Compute a line-level unified diff of a living-doc version (:id, the target) against a base version. Use for the version-diff view. Base defaults to the immediate predecessor; pass ?base=<version_id> for a specific version or ?base=current for the promoted current. Returns hunks (context + add/remove lines, word-level segments on edits), +/- stats, a truncated flag, and base_missing=true for v1. Returns null when the target is missing or cross-business.

• HTTP: GET /blueprint-outputs/:id/diff
• Required scopes: workflows:read
• Required role: viewer
• Stability: beta
• Version: 1.0

Input

Field	Type	Required	Description
id	string	yes
base	string	no

blueprint_output.get

Get Blueprint Output

Fetch a single blueprint living-doc version by id, including promotion status, the kb_doc_id it serves, and provenance (source_run_id, as_of, diff_summary, quality_score, promoted_by/at). Use when rendering a version detail or the promote confirmation. Returns null when the version does not exist or belongs to a different business.

• HTTP: GET /blueprint-outputs/:id
• Required scopes: workflows:read
• Required role: viewer
• Stability: beta
• Version: 1.0

Input

Field	Type	Required	Description
id	string	yes

blueprint_output.list

List Blueprint Outputs

List a recurring blueprint's living-doc versions for this business, newest first. Use when rendering the Versions tab, finding the current canonical (each row carries is_canonical), or auditing what was produced and when. Filter by blueprint_id. Returns lean version rows; fetch blueprint_output://<id> for full provenance.

• HTTP: GET /blueprint-outputs
• Required scopes: workflows:read
• Required role: viewer
• Stability: beta
• Version: 1.0

Input

Field	Type	Required	Description
blueprint_id	string	no
limit	integer	no
offset	integer	no

blueprint_output.list_comments

List Blueprint Output Version Comments

List a workflow living-doc version’s review threads for the Workflows Versions tab (:id is the output / knowledge_bank_documents id). Use when a reviewer opens a version diff and wants the agent’s per-edit rationale plus teammate replies as resolvable threads, with the persisted block anchors. The workflows:read counterpart of knowledge_base://<id>/comments; identical data + ACL.

• HTTP: GET /blueprint-outputs/:id/comments
• Required scopes: workflows:read
• Required role: viewer
• Stability: beta
• Version: 1.0

Input

Field	Type	Required	Description
id	string	yes

blueprint_run.get

Get Blueprint Run

Fetch a single blueprint run by id, including step_states, current_step_id, tokens used, and any error message. Use when rendering an individual run timeline or auditing a finished run. Returns null when the run does not exist or belongs to a different business.

• HTTP: GET /blueprint-runs/:id
• Required scopes: workflows:read
• Required role: viewer
• Stability: stable
• Version: 1.0

Input

Field	Type	Required	Description
id	string	yes

blueprint_run.list

List Blueprint Runs

List blueprint runs (one per agent_blueprint invocation) for this business, newest first. Use when rendering the run-history timeline, filtering by blueprint_id, or auditing past runs by status (queued|running|complete|error|canceled). Returns lean rows; fetch blueprint_run://<id> for the full step_states + tokens.

• HTTP: GET /blueprint-runs
• Required scopes: workflows:read
• Required role: viewer
• Stability: stable
• Version: 1.0

Input

Field	Type	Required	Description
blueprint_id	string	no
status	string	no
limit	integer	no
offset	integer	no

blueprint_schedule.create

Create Blueprint Schedule

Create a recurring CRON schedule that fires a saved agent_blueprint headlessly. An agent can call this to schedule unattended runs (e.g. "run the research-report blueprint every Monday at 9am"). Provide the blueprint id, a cron expression, optional timezone (default UTC), inputs, and name; the platform validates the blueprint + cron, stores the schedule, and fires a headless run on each tick. Returns the created schedule (manage existing ones via blueprint_schedule.list/get/update/delete).

• HTTP: POST /agent-blueprints/:id/schedules
• Required scopes: workflows:delegate
• Required role: editor
• Stability: stable
• Version: 1.0
• Tags: blueprints, schedules, write, workflow

Input

Field	Type	Required	Description
id	string	yes	Starter slug / UUID or tenant agent_blueprint artifact id to schedule.
cron	string	yes	Cron expression (5- or 6-field, e.g. "0 9 * * 1" for 9am every Monday). Validated before the schedule is stored.
timezone	string	no	IANA timezone the cron is evaluated in (e.g. "America/New_York"). Defaults to UTC.
inputs	object	no	Resolved input values keyed by the blueprint's declared input names. Declared defaults fill the rest at fire time.
name	string	no	Optional human label for the schedule.
enabled	boolean	no	Whether the schedule is active. Defaults to true.

blueprint_schedule.delete

Delete Blueprint Schedule

Delete a blueprint cron schedule by id. An agent can call this to remove a scheduled run entirely (to merely pause it, use blueprint_schedule.update with enabled:false). Returns success, or not_found if it does not exist in this business.

• HTTP: DELETE /blueprint-schedules/:id
• Required scopes: workflows:delegate
• Required role: editor
• Stability: stable
• Version: 1.0
• Tags: blueprints, schedules, write, workflow

Input

Field	Type	Required	Description
id	string	yes	Schedule id to delete.

blueprint_schedule.get

Get Blueprint Schedule

Fetch a single blueprint cron schedule by id, including its cron, timezone, inputs, enabled flag, and last/next run bookkeeping. Use when rendering or auditing one schedule. Returns null when the schedule does not exist or belongs to a different business.

• HTTP: GET /blueprint-schedules/:id
• Required scopes: workflows:read
• Required role: viewer
• Stability: stable
• Version: 1.0

Input

Field	Type	Required	Description
id	string	yes

blueprint_schedule.list

List Blueprint Schedules

List cron schedules (one per scheduled agent_blueprint) for this business, newest first. Use when rendering the schedules surface, filtering by blueprint_id, or showing only enabled/disabled schedules. Returns the schedule rows plus total/limit/offset.

• HTTP: GET /blueprint-schedules
• Required scopes: workflows:read
• Required role: viewer
• Stability: stable
• Version: 1.0

Input

Field	Type	Required	Description
blueprint_id	string	no
enabled	boolean	no
limit	integer	no
offset	integer	no

blueprint_schedule.update

Update Blueprint Schedule

Update an existing blueprint cron schedule: change its cron, timezone, inputs, name, or enable/disable it. An agent can call this to retune or pause a scheduled run. A cron/timezone change is re-validated before it is saved. Returns the updated schedule, or not_found if it does not exist in this business.

• HTTP: PATCH /blueprint-schedules/:id
• Required scopes: workflows:delegate
• Required role: editor
• Stability: stable
• Version: 1.0
• Tags: blueprints, schedules, write, workflow

Input

Field	Type	Required	Description
id	string	yes	Schedule id to update.
cron	string	no	New cron expression (validated).
timezone	string	no	New IANA timezone.
inputs	object	no	New input values keyed by the blueprint's input names.
name	`string	null`	no
enabled	boolean	no	Enable (true) or disable (false) the schedule.

blueprints.create

Create Blueprint

Author and save a brand-new agent_blueprint recipe for this workspace from a v1 DSL body. Use when a user composes a fresh automation in the editor (or asks chat to draft one) and wants it persisted. The body runs through the full authoring moat — schema, step-id uniqueness, reference resolution, sub-blueprint cycle detection, tool-allowlist format — before any row is written; failures return a structured validation_failed list. Starters are read-only; fork one to base a recipe on it.

• HTTP: POST /agent-blueprints
• Required scopes: versioning:write
• Required role: editor
• Stability: stable
• Version: 1.0
• Tags: blueprints, write

Input

Field	Type	Required	Description
content	object	yes	The v1 agent_blueprint DSL body. Read step_kind://list for the grammar.
title	string	no
description	string	no
status	string	no
tags	array<string>	no
metadata	object	no

blueprints.delete

Delete Blueprint

Archive a workspace blueprint so it drops out of the active library while staying recoverable. Use when a user removes a recipe they no longer need; the row is soft-deleted (hidden, reversible) rather than destroyed. Reverse it later with the unarchive op. Code-shipped starters cannot be archived; permanent removal of a tenant blueprint goes through the generic artifact delete path.

• HTTP: DELETE /agent-blueprints/:id
• Required scopes: versioning:write
• Required role: editor
• Stability: stable
• Version: 1.0
• Tags: blueprints, write

Input

Field	Type	Required	Description
id	string	yes	Artifact UUID of the tenant blueprint.

blueprints.promote_output

Promote Blueprint Output

Promote a blueprint version to canonical — the single served document a recurring blueprint's living doc resolves to. Use when a human reviews a proposed version and publishes it: the prior canonical is superseded and archived out of retrieval, the chosen version becomes the citable doc, and the event is stamped with who/when. Name the target by output_id (a registered version) or kb_doc_id (a raw/out-of-band doc, registered on the fly). Promotion is human-gated by design.

• HTTP: POST /agent-blueprints/:id/promote
• Required scopes: versioning:write
• Required role: editor
• Stability: beta
• Version: 1.0
• Tags: blueprints, living-doc, write

Input

Field	Type	Required	Description
id	string	yes	Starter slug / UUID or tenant agent_blueprint artifact id to set the canonical for.
output_id	string	no	The blueprint_outputs row id (registered version) to promote.
kb_doc_id	string	no	A knowledge_bank_documents id to promote directly (registered as a candidate first if needed).

blueprints.unarchive

Unarchive Blueprint

Bring a previously archived workspace blueprint back into the active library. Use when a user undoes a blueprint removal and wants the recipe usable again. Operates only on tenant-owned rows that are currently archived; code-shipped starters have no archive state to restore.

• HTTP: POST /agent-blueprints/:id/unarchive
• Required scopes: versioning:write
• Required role: editor
• Stability: stable
• Version: 1.0
• Tags: blueprints, write

Input

Field	Type	Required	Description
id	string	yes	Artifact UUID of the archived tenant blueprint.

blueprints.update

Update Blueprint

Save edits to a workspace blueprint the caller already owns. Use when a user modifies an existing recipe and persists the change. The content field is a full replacement body re-checked through the authoring moat; title, description, status, tags, and metadata patch the surface row. Code-shipped starters reject — fork them first. A slug collision returns code slug_conflict with the conflicting artifact id for inline field UX.

• HTTP: PATCH /agent-blueprints/:id
• Required scopes: versioning:write
• Required role: editor
• Stability: stable
• Version: 1.0
• Tags: blueprints, write

Input

Field	Type	Required	Description
id	string	yes	Artifact UUID of the tenant blueprint.
content	object	no	Full replacement v1 DSL body. Omit to patch surface fields only.
title	string	no
description	string	no
status	string	no
tags	array<string>	no
metadata	object	no
change_summary	string	no

blueprints.validate

Validate Blueprint

Dry-run a blueprint DSL body through the full authoring moat without persisting anything. Use when an editor wants author-time feedback on a draft before saving — the response lists schema errors, duplicate step ids, unresolvable references, sub-blueprint cycles, and tool-allowlist format problems, with an overall valid flag. Nothing is written to the workspace.

• HTTP: POST /agent-blueprints/validate
• Required scopes: versioning:write
• Required role: editor
• Stability: stable
• Version: 1.0
• Tags: blueprints

Input

Field	Type	Required	Description
content	object	yes	Candidate v1 agent_blueprint DSL body to validate.

external_search_jobs.read_stream

External Search Job Stream

Read when discovering the live SSE event channel for a running external_search job spawned with subscribe_to_events=true. Returns a small JSON descriptor with the canonical channel URI, the companion HTTP stream path, the Redis pub/sub channel name, and content-type=text/event-stream so callers can open an SSE EventSource without re-deriving the path. Streaming itself happens on the GET HTTP route.

• HTTP: GET /external-search-jobs/:jobId/stream
• Required scopes: external_search:execute
• Required role: viewer
• Stability: beta
• Version: 1.0

Input

Field	Type	Required	Description
jobId	string	yes	The external_search job id (UUID) returned by external_search.execute when called with subscribe_to_events=true.

finish

Finish Turn

Signal that the task is complete. Call this as the final tool when you have finished the work. Arguments may include an optional summary and an optional artifact_id produced in a prior turn. Use when the full job is done, not mid-task.

• HTTP: (not exposed as REST)
• Required scopes: workflows:complete
• Required role: viewer
• Stability: stable
• Version: 1.0
• Tags: system, terminal

Input

Field	Type	Required	Description
summary	string	no	Short summary of the work completed.
artifact_id	string	no	Optional id of the platform artifact that carries the deliverable.

knowledge_base.make_living

Make Knowledge Base Document Living

Convert a static knowledge base document into a self-refreshing living doc by synthesizing a refresh workflow (an agent_blueprint) bound to the doc version group and optionally scheduling it. Use when a user wants a KB doc to auto-regenerate on a cadence instead of going stale. Authors the recipe, forces the upload step onto the doc group, validates, persists the blueprint, and creates a cron schedule when a cron is supplied.

• HTTP: POST /knowledge_base/documents/:id/make-living
• Required scopes: workflows:write, knowledge_base:write
• Required role: editor
• Stability: stable
• Version: 1.0
• Tags: knowledge-base, blueprints, write

Input

Field	Type	Required	Description
id	string	yes	UUID of the knowledge base document to make living.
schedule_cron	string	no	Optional 5-field cron expression. When provided, the refresh blueprint is scheduled to run on this cadence; omit to create the blueprint without a schedule.
schedule_timezone	string	no	Optional IANA timezone for the schedule (e.g. "America/New_York"). Defaults to UTC. Ignored when schedule_cron is omitted.

prompt.get

Get Prompt Fragment

Fetch a single platform prompt fragment by id (full body included). Use when a blueprint llm step's prompt_resources references prompt://<scheme>/<id>; the runner or canvas inlines the returned body verbatim. Path is the slash-namespaced fragment id (e.g. content_writer/grounding_rules). Returns null when the id is not registered.

• HTTP: GET /prompt-fragments/:id
• Required scopes: artifacts:read
• Required role: viewer
• Stability: stable
• Version: 1.0

Input

Field	Type	Required	Description
id	string	yes

prompt.list

List Prompt Fragments

List every registered platform prompt fragment (named, version-controlled chunks of expertise that agent_blueprint llm steps can compose via prompt_resources). Use when authoring or reviewing a blueprint to discover what reusable prompt patterns exist; fetch a single fragment via prompt://<scheme>/<id> for the full body. Optional scheme filter (e.g. ?scheme=content_writer).

• HTTP: GET /prompt-fragments
• Required scopes: artifacts:read
• Required role: viewer
• Stability: stable
• Version: 1.0

Input

Field	Type	Required	Description
scheme	string	no

retrieve_scratchpad

Retrieve Scratchpad Slot

Read a slot from the session scratchpad. Scratchpad slots are populated by prior tool calls when the output was too large to inline. Use when you need a full truncated tool result that was stored under a key you received earlier.

• HTTP: (not exposed as REST)
• Required scopes: workflows:read
• Required role: viewer
• Stability: stable
• Version: 1.0
• Tags: system, session-state

Input

Field	Type	Required	Description
key	string	yes	The scratchpad slot key to read. Keys are returned in prior tool results.

step_kind.get

Get Step Kind

Fetch metadata for a single step kind by id (tool / llm / loop / branch / parallel / blueprint / transform / assert). Use when rendering a single canvas node or fetching catalog details for a specific step. Returns null when the id is not registered.

• HTTP: GET /step-kinds/:id
• Required scopes: artifacts:read
• Required role: viewer
• Stability: stable
• Version: 1.0

Input

Field	Type	Required	Description
id	string	yes

step_kind.list

List Step Kinds

List every step kind the agent_blueprint DSL supports (tool, llm, loop, branch, parallel, blueprint, transform, assert). Use when authoring or visualizing a blueprint to discover the available control-flow + data + agent_call primitives. Returns id, name, description, category for each kind.

• HTTP: GET /step-kinds
• Required scopes: artifacts:read
• Required role: viewer
• Stability: stable
• Version: 1.0

Input

No structured input.

trigger_kind.get

Get Trigger Kind

Fetch metadata for a single trigger kind by id (manual / schedule / event / webhook). Use when rendering a single trigger configuration card. Returns null when the id is cataloged but its handler is not yet implemented (e.g. webhook).

• HTTP: GET /trigger-kinds/:id
• Required scopes: artifacts:read
• Required role: viewer
• Stability: stable
• Version: 1.0

Input

Field	Type	Required	Description
id	string	yes

trigger_kind.list

List Trigger Kinds

List every trigger kind the agent_blueprint DSL supports (manual today; schedule/event/webhook in later waves). Use when authoring or visualizing a blueprint to discover how it can be invoked. Returns id, name, description, category, and the wave each kind ships.

• HTTP: GET /trigger-kinds
• Required scopes: artifacts:read
• Required role: viewer
• Stability: stable
• Version: 1.0

Input

No structured input.