Automation your team actually owns
GoRunner is a workflow automation platform in the class of n8n and Zapier — a visual builder, an AI copilot that drafts whole flows, durable execution, and a real API behind every automation you ship.
# trigger a production run and wait for the result curl -X POST "$BASE/api/workflows/$ID/run?environment=production&wait=20" \ -H "Authorization: Bearer grt_…" \ -d '{"orderId": 9137}' {"data": {"status": "completed", "steps": [...] }}
01 · Run it
Get it running
Your first workflow in about five minutes — then the workflow model, triggers, runs, publishing, and Keel.
02 · Build on it
API Reference
Every endpoint its own page, from the machine-checked OpenAPI contract — try-it playground and code in seven languages.
03 · Automate it
Connect an agent
Wire Claude or any MCP client to your workspace: agents that build workflows, and workflows that become the agent’s tools.
Why teams switch
Built for the parts the giants get wrong
A canvas that tells the truth
Branches, loops, per-step retries and error paths on a recursive tree — with test runs that show every step’s real input and output before anything ships.
Keel — the AI builder
Describe the automation in plain words. Keel plans, resolves against the piece catalog, drafts the whole flow in two model passes, and pre-flies the wiring before you ever click Apply.
Draft vs. production, enforced
Editing a live workflow never changes what production runs. Publish is an explicit, versioned, validated step — and restore is one click.
Durable execution
Runs survive restarts and deploys, pause for approvals or delays without holding a worker, and resume days later. Every step is recorded.
A real public API
Scoped tokens, one error envelope, idempotency keys, validate-before-write — the whole editor surface as HTTP, under a machine-checked OpenAPI contract.
Agent-native by design
Two MCP servers: agents can build workflows in your workspace, and your published workflows become tools an agent calls by name.