Automate the checks. Keep the decision.

Have an agent prepare a product update, pause for a human decision, then send the approved result to your API.

Get the YAML ↓Workflow docs

THE FLOW

From trigger to outcome.

A repeatable process with a visible human decision before the protected action.

01 →
Prepare
Run an agent with a pinned binding revision and workspace to prepare a draft.
02 →
Wait
Pause at a wait step. Review the draft in the run before approving or rejecting it.
03 →
Continue
Run the approved next step and preserve the decision with the run.

USE THIS RECIPE

Your workflow,
in YAML.

Replace the agent binding ID, revision ID, revision digest, and workspace ID with your configured values.
Use the exact sha256: digest for that binding revision, including all 64 hexadecimal characters.
Set publish_url and create API_TOKEN for the destination’s X-API-Key header. The endpoint receives the full agent result as JSON.

Configure secret references →

Copy or download the manifest, configure your values, and add it through Workflows or your AI workspace. Opening the console does not create or enable a workflow.

Run AI agents
Wait for approval
Call APIs

review-and-publish.yaml

Download

workflow_id: review-and-publish
title: Review before publishing
enabled: true

context:
  agent_binding_id: YOUR_AGENT_BINDING_ID
  agent_revision_id: YOUR_AGENT_REVISION_ID
  agent_revision_digest: "sha256:YOUR_REVISION_DIGEST"
  workspace_id: YOUR_WORKSPACE_ID
  publish_url: https://api.example.com/drafts

steps:
  - step_id: draft
    kind: agent
    command:
      agent_binding_id: "{{ .context.agent_binding_id }}"
      agent_binding_revision_id: "{{ .context.agent_revision_id }}"
      agent_binding_revision_digest: "{{ .context.agent_revision_digest }}"
      workspace_id: "{{ .context.workspace_id }}"
      input: Draft a short product update for review.

  - step_id: approve
    kind: wait
    dependencies: [draft]

  - step_id: publish
    kind: http
    dependencies: [approve]
    command:
      method: POST
      url: "{{ .context.publish_url }}"
      headers:
        X-API-Key: secret://API_TOKEN
      body: "{{ .steps.draft.output }}"

3 stepsOpen Workflows

Setup Replace the agent binding, revision, SHA-256 digest, workspace, and publish URL with your configured values. Add API_TOKEN and approve the draft in Adios before the final step runs.

EXPECTED RESULT

Know what to look for.

The run waits after the agent completes. Approval releases the publish step, which posts the draft step’s JSON output. Rejection stops the run; an unanswered request remains waiting.

Approval is an action inside a run, not a trigger. This example starts manually. Its final HTTP step sends the complete agent result; adapt the destination or add a data-json step if you need one field.

TEST YOUR SETUP

Check a complete run.

Start manually and inspect the draft step output. The publish step must stay pending while approval is unanswered.
Approve one test run and reject another. Approval allows publishing; rejection fails the wait step and stops the run.
Choose who will review runs in your team. The wait action has no YAML fields for reviewer assignment or automatic expiry.

Adapt this recipe with your AI agent

Help me design an approval workflow in Adios. Ask what action requires approval, who should review it, and what evidence they need. Prepare the evidence, then use the platform's approval/wait controls before the protected step. Define rejection and unanswered-request behavior explicitly. Keep the approved action tied to the reviewed inputs. Show me the workflow and tests first; do not approve, deploy, or enable any production action on my behalf.

Open Workflows ↗

KEEP CONNECTING