list_, get_, create_, update_, delete_, plus one tool per skill (skill_<slug>) and per workflow (workflow_<slug>). Each tool does the same thing as the API call in its last column, so the API page has the full fields.
For how to connect, see Connect your AI assistant. For what each tool changes and costs, see Limits and side effects.
Skills
Over OAuth,
skills:invoke lets an assistant run skills and also create and improve them (create_skill, get_skill_creation, improve_skill, get_skill_improvement). Changing, deleting, and scheduling skills need skills:write, which you tick on the consent screen under Optional extra access (“Create, edit and delete your skills”). It is never granted by default.
One tool per skill: skill_<slug>
Every skill you can run is its own tool, named skill_ plus its slug. A skill with the slug austin-dentists becomes skill_austin-dentists. The list holds skills you own and skills shared into your workspaces. Marketplace skills appear once you clone them.
Each tool has:
- A description: what the skill does, plus a short track record (runs, recent success rate, typical time and row count, the last error, and its fast path if it has one).
- An input schema: the skill’s own inputs, plus optional
max_results(0–500,0for the whole list) andversion. - An output schema: the row shape, so the assistant can rely on field names.
skills:read. Calling needs skills:invoke. It works like Run a skill and wait. The list is rebuilt each time the assistant refreshes it, so a new skill shows up on the next refresh.
invoke_skill (skill_id, optional params, max_results) runs a skill by id. Prefer the skill_<slug> tools: they carry typed schemas.
Runs
Older tools that still work:
get_run_status (run_id, a workflow run) and list_recent_runs (optional limit, your latest workflow runs). New assistants should use get_run and list_workflow_runs.
Improvements
When a skill run comes back short, its result has a
diagnosis with hints the assistant can pass as feedback. See Improve a skill.
Schedules
A skill or workflow has one schedule. Skill schedule ids start withsklsch_, workflow ones with wfsch_.
Workflows
workflows:write is on API keys. For an assistant it is opt-in: tick “Create, edit and delete your workflows and schedules” under Optional extra access on the consent screen.
One tool per workflow: workflow_<slug>
Every workflow you can run is its own tool, for example workflow_competitor-prices. Its input schema is the workflow’s inputs and its output schema is the run. Calling it needs workflows:invoke. It works like Run a workflow and wait. Clients that support the MCP Tasks extension get a task they can poll instead of waiting.
invoke_workflow (workflow_id, optional inputs) starts a workflow by id and returns a run_id. Prefer the workflow_<slug> tools.
Each workflow is also a resource at valendata://workflows/{workflow_id}/latest. Reading it returns the rows of its last finished run, at no cost. Needs workflows:read.
Account
account:read is on API keys. For an assistant it is opt-in: tick “See your credits balance and usage” under Optional extra access on the consent screen.