Automation platforms

Greenlight is a plain HTTP API. Any workflow tool that can send REST requests and receive webhooks can act as the “agent” — no custom Greenlight plugin required.

Use automation platforms when you want visual workflows, scheduled jobs, or glue between Greenlight and the rest of your stack (CI, ticketing, databases, LLM APIs) without writing a long-lived service.

How it fits

Outbound (automation → Greenlight) — POST /v1/channels/new and POST /v1/prompts/new create durable resources; POST /v1/messages/send delivers chat text on MESSAGE channels. Authenticate with X-API-Key when USE_AUTH=true.

Inbound (Greenlight → automation) — prompt answers hit the prompt’s callback_url; MESSAGE channels post message.created to the channel’s callback_url. See Callbacks for payload shapes.

Recommended platforms

Greenlight does not endorse or maintain connectors for these tools. The Agent API is stable HTTP — pick the platform your team already runs.

Typical patterns

Deploy approval (PROMPT)

1. Create channel once — POST /v1/channels/new with channel_type: "PROMPT" (no channel-level callback_url; each prompt supplies its own).
2. Trigger — GitHub Actions, GitLab CI, Jenkins, or a schedule fires your workflow.
3. Create prompt — POST /v1/prompts/new with options, correlation_id (build ID), and a webhook URL from your automation tool.
4. Wait for human — Greenlight delivers the card to Slack/Telegram/etc.
5. Resume workflow — On signed callback, branch on answer.value (Approve / Reject) and continue the pipeline or open a ticket.

Many workflows can share one PROMPT channel (for example a single #ops-approvals chat); each prompt still resumes at its own callback_url. See Features & benefits.

Ops chat bot (MESSAGE)

1. Create once — POST /v1/channels/new with channel_type: "MESSAGE" and your automation tool’s webhook URL as callback_url (where Greenlight forwards inbound chat as unsigned message.created events).
2. Configure the platform — point the chat platform at Greenlight’s inbound webhook: {PUBLIC_WEBHOOK_URL}/webhooks/{organization_id}/{platform}/{channel_id} (see Platforms). This is not the same URL as callback_url.
3. On message — Handle message.created at your callback_url; call an LLM API, lookup a runbook, or update a ticket.
4. Reply — POST /v1/messages/send back through Greenlight to the same chat.

See Concepts for PROMPT vs MESSAGE diagrams.

Wiring checklist

Base URL: http://localhost:8100 locally, or your PUBLIC_WEBHOOK_URL origin in production. Full endpoint list: Agent API.

Prompt callback signing in no-code tools

Prompt answer callbacks are signed with CALLBACK_SIGNING_SECRET:

X-Signature: sha256=<hmac-sha256-hex-of-raw-body>

Most visual builders do not verify HMAC out of the box. Options:

• Code step — n8n Code node, Pipedream Node.js, Node-RED Function node, or Power Automate inline script. Use the raw request body; re-serialized JSON will break the hash.
• Trusted network — automation webhook only reachable on a private network or VPN (weaker than signature verification).
• Poll instead — skip webhooks and poll GET /v1/prompts/{id} from the workflow (simpler, higher latency).

MESSAGE message.created events are unsigned; protect the webhook URL (random path, auth header on the automation side, IP allowlist).

Minimal HTTP examples

Prefix paths with your API origin (http://localhost:8100 locally, or your PUBLIC_WEBHOOK_URL in production). Send X-API-Key when USE_AUTH=true.

Create a PROMPT channel (one-time per target chat):

POST /v1/channels/new
X-API-Key: <GREENLIGHT_API_KEY>
Content-Type: application/json
 
{
  "channel_id": "ops-approvals",
  "platform": "slack",
  "target_chat_id": "C01234567",
  "credentials": { "bot_token": "xoxb-...", "signing_secret": "..." },
  "channel_type": "PROMPT"
}

Response 201:

{ "status": "Channel ops-approvals registered." }

Create a prompt (repeat per approval):

POST /v1/prompts/new
X-API-Key: <GREENLIGHT_API_KEY>
Content-Type: application/json
 
{
  "channel_id": "ops-approvals",
  "text": "Approve production deploy?",
  "options": ["Approve", "Reject"],
  "correlation_id": "build-9182",
  "callback_url": "https://hooks.example.com/greenlight/answers"
}

Create a MESSAGE channel (one-time per target chat):

callback_url is where Greenlight forwards inbound chat to your automation. Configure the chat platform separately to post events into Greenlight at {PUBLIC_WEBHOOK_URL}/webhooks/{organization_id}/{platform}/{channel_id} — see Platforms.

POST /v1/channels/new
X-API-Key: <GREENLIGHT_API_KEY>
Content-Type: application/json
 
{
  "channel_id": "ops-slack-chat",
  "platform": "slack",
  "target_chat_id": "C01234567",
  "credentials": { "bot_token": "xoxb-...", "signing_secret": "..." },
  "callback_url": "https://hooks.example.com/greenlight/messages",
  "channel_type": "MESSAGE"
}

Reply in chat on that MESSAGE channel:

POST /v1/messages/send
X-API-Key: <GREENLIGHT_API_KEY>
Content-Type: application/json
 
{
  "channel_id": "ops-slack-chat",
  "text": "Deploy is live in prod."
}

POST /v1/channels/new returns 409 if channel_id already exists — use PATCH /v1/channels/{id} to update or reactivate a deactivated channel.

Expose local automation webhooks with ngrok  or Cloudflare Tunnel  when Greenlight runs on your machine.

When to use code instead

Automation platforms excel at triggers, branching, and integrations. Prefer a small service (or Agent integration) when you need:

• High-volume MESSAGE traffic with strict latency
• Complex multi-prompt state machines
• Centralized signature verification and idempotency

Next

• Agent integration — same HTTP model for custom agents
• Callbacks — payload fields and verification
• Prompts — options, TTL, correlation_id
• Channels — create channel schema