Docs

Schedules & workflows

Run tools on a cron schedule, and chain tool calls without a model in between.

Scheduled calls (cron)

The built-in cron connector calls a tool of any connector on a schedule, as a profile. The server keeps the schedules and does the calling; the connector and the web UI's Schedules section only manage them.

callTool
cron.tools schedule -b '{"connector":"time","tool":"now","schedule":"*/15 * * * *","input":{"timeZone":"Europe/Berlin"}}'
cron.tools listSchedules -b '{"connector":"time"}'
cron.tools listRuns -b '{"status":"failed","limit":10}'
  • Expressions are standard five-field cron or @hourly, @daily, @weekly, @monthly, @yearly, in UTC or an IANA time zone per job.
  • A job runs as its profile: it can use only the connectors and tools that profile exposes, with its sign-ins.
  • Shortest interval is 5 minutes by default (MESHER_CRON_MIN_INTERVAL_MINUTES).
  • Limits: 48 schedules per profile, 240 per user by default.
  • Runs record input, output and status (succeeded, failed, skipped) and are kept for a day by default (MESHER_CRON_RESULT_TTL_MINUTES). A run missed while the server was down is not made up.
  • Servers sharing an S3 bucket never both run a slot.

Workflows

A workflow chains the tool calls of other connectors without a model in between. It is a small Rust program of step functions compiled to wasm; the manifest names each step and where its ok and error paths may lead. A step gets what the last call returned and answers with the calls to make next, in parallel if need be. It has no network: the host makes the calls as the profile that runs it.

callTool
wasmrs.tools newProject -b '{"name":"digest","template":"workflow"}'
digest.tools startRun -b '{"input":{"city":"Berlin"}}'
digest.tools getRun -b '{"runId":"0k6832IujQX"}'
digest.tools listRuns -b '{"status":"running"}'
cron.tools schedule -b '{"connector":"digest","tool":"startRun","schedule":"0 9 * * mon-fri"}'
  • startRun answers with a run id at once. A run stops after an hour, 100 steps or 16 MiB in one payload.
  • Retries: "retries": 2 on a step (0 to 3) retries a failed call after 1 s, 2 s, then 4 s, before the error goes to its catch.
  • Waiting: a step can park its branch until an event arrives, with a timeout, or sleep for a while. Send events with sendEvent, the API, or the run page.
  • Limits: 16 running runs per profile and 64 per user by default.
  • The web UI lists workflows under Workflows: a graph, the steps, and each run with the step every branch is at.

The workflows built-in lists, starts, reads, stops and feeds the runs of the calling profile:

callTool
workflows.tools listWorkflows -b '{}'
workflows.tools startRun -b '{"workflow":"digest","input":{"city":"Berlin"}}'
workflows.tools cancelRun -b '{"runId":"0k6832IujQX"}'

To write one, follow the workflows chapter of the authoring guide.

Last updated November 21, 2026 · This page as Markdown · Full reference on GitHub