Thanks to visit codestin.com
Credit goes to docs.sim.ai

Concept

Overview

A deployment is a published version of a workflow that outside callers can run. While you build, a workflow lives on your canvas as a draft only you can run. Deploying publishes a fixed copy of that draft and gives it an address: a REST endpoint, a chat page, a set of MCP tools, or an A2A agent. Each is a surface, a channel callers reach the same published workflow through.

Like publishing a book, building and deploying are separate acts. You draft and revise freely on the canvas. When you click Deploy, Sim prints a fixed edition. Readers get that edition, not your latest draft, until you publish a new one. You can always go back to an earlier printing.

The deployment model

Snapshot

A snapshot is an immutable copy of your workflow taken at the moment you deploy. It is frozen: editing the canvas afterward changes your draft, not the snapshot. Every call to your deployment runs against the snapshot, so a half-finished edit on the canvas never reaches a live caller. Publishing again takes a fresh snapshot of the current draft.

Version

Each snapshot is recorded as a numbered version (v1, v2, and so on) in the Versions table, tagged with who deployed it and when. Versions are independent: one is live at a time, marked with a green dot, and the rest stay available to rename, describe, load back to the canvas, or promote. You publish v1 with Deploy and every later Update adds the next version.

Live version

The live version is the one snapshot currently exposed through every surface. The General tab shows it as Live Workflow, a read-only minimap of exactly what callers run. Only one version is live at a time. Promoting a different version, or updating to a new one, swaps which snapshot is live without affecting the others.

Update

After you change the canvas on a deployed workflow, an Update deployment badge appears in the toolbar to flag that your live version is behind your draft. Clicking Update takes a new snapshot, records it as the next version, and makes it live, all without reopening the deploy modal.

Promote to live

Promote to live makes an earlier version the live one again. It is the fast path for rollback: if a new deployment misbehaves, promote the last known-good version to restore it instantly. Promoting reuses an existing snapshot, so it does not create a new version.

Editing the canvas never changes what is live. The live snapshot only moves when you click Update (publish a new version) or Promote to live (restore an earlier one), and both of those are deliberate actions you take.

The surfaces

Every surface runs the same live snapshot. You manage them from the tabs of the deploy modal.

The most common surface is the API. Once deployed, your workflow answers at:

POST https://sim.ai/api/workflows/{workflow-id}/execute

Send an object in the request body matching the workflow's Input Format. The response is an object holding the block outputs and execution metadata.

curl -X POST https://sim.ai/api/workflows/{workflow-id}/execute \
  -H "X-API-Key: $SIM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "input": "Refund request from customer #4821" }'

Versioning in practice

Treat the live version as production and the canvas as staging. Run the workflow on the canvas until it behaves, then Deploy or Update to publish. Callers keep hitting the previous snapshot until the new version goes live, so there is no half-deployed state. If a new version causes trouble, Promote to live the previous one to roll back at once, then fix the draft and update again when it is ready. The logs record every run against every version, so you can confirm which snapshot a given execution used.

Next

On this page