Operating model

Define the responsibility, approve its tools, prove one run, then let the schedule own it.

Concepts

How Work on Repeat works Organizations and projects Routines Runs and traces Run lifecycle Connections Tool permissions Models

Working well

Writing instructions Variables Schedules and timezones Security model

Answers

Troubleshooting FAQ Glossary

Knowledge baseAnswers

Troubleshooting

Every failure you are likely to meet, what causes it, and what to change.

Start with the run's error category. It names the layer that failed, which is usually enough to know what to change.

Runs

Scheduling

Webhooks

API

Still stuck

Every API response and every run carries a requestId. Include it — it identifies the exact request or execution in the service logs.

Security model

Project isolation, secret handling, bounded execution, explicit tool permissions, and outbound network rules.

FAQ

Model choice, write access, failure behaviour, cost, and what Work on Repeat is not.

Troubleshooting

Runs

step_limit — the run stopped without finishing

timeout / timed_out — the run exceeded its deadline

tool_denied — the model asked for something it may not use

tool_error — an approved tool ran and failed

mcp_discovery — discovery failed

mcp_transport — a call failed mid-run

gateway_rate_limit / gateway_timeout / gateway_error

invalid_model — the model is not in the directory

authentication / configuration

cancelled — the run was aborted

internal — an unexpected failure

Scheduling

Activate is unavailable

The routine went back to draft on its own

It ran at the wrong hour

It has not run at all since activation

invalid_schedule when saving

Webhooks

Signature verification fails

The same event arrives more than once

Deliveries stopped after four tries

Nothing arrives at all

API

401 unauthorized

403 insufficient_scope

403 session_required

404 on a resource you can see in the dashboard

429 rate_limit_exceeded

413 body_too_large

Still stuck

On this page