Live chat API and developer platform

The live chat API that fits your stack

A REST API with OpenAPI docs, signed webhooks that retry, a JavaScript widget API, a CLI and a hosted MCP server. Wire Pengo into your product, your backend and the AI tools your team already uses.

Widget API on every planSigned webhooks from GrowthREST API from Growth

Animation: Tallybook's developer view of one support chat. Ruth Adebayo starts a chat about a bank feed, and a signed chat.created webhook is delivered to Tallybook's endpoint, which answers 200. In a terminal, the Pengo CLI marks the conversation high priority with a bank-feeds tag and sends a reply. Then an AI client connected to Pengo's hosted MCP server lists the open conversations, reads Ruth's and adds an internal note for the team.

  1. How it works
  2. How it works
  3. How it works

Does Pengo have a live chat API?

Yes. Pengo has a REST API for conversations, messages, contacts, visitors, tags, knowledge and audit logs, documented with OpenAPI, plus signed webhooks, a JavaScript widget API, a CLI and a hosted MCP server. Developers use it to sync conversations into their own systems, reply from code and give AI agents safe access to the inbox.

REST API
v1 with scoped API keys, from Growth
Webhooks
Signed, retried and dead-lettered, from Growth
Widget API
JavaScript methods and events, on every plan
AI agents
Hosted MCP server with OAuth 2.1
01 The platform

Nine ways to make Pengo part of your product

From a one-line widget call to an AI agent that queries your backend, every surface is documented, scoped and signed.

Growth+

REST API v1

Conversations, messages and notes, contacts, visitors, agents, tags, knowledge and audit logs. Every endpoint is documented in the OpenAPI reference.

Growth+

API keys

Create a key per integration in Settings, send it as a Bearer token, revoke it anytime.

Growth+

Signed, retried webhooks

HMAC-signed with a timestamp, retried with backoff, and dead-lettered if your endpoint stays down.

Scale+

Workflows

Trigger → conditions → waits → actions, with AI steps. Let Intelligence classify the chat, wait up to 30 days, then tag, assign, reply, post to Slack or call a webhook.

Every plan

Widget JS API

18 methods, from open, toggle and identify to setLocale, getUnread, setVisitor, track and showMessage, plus events. Calls queue until the widget loads.

Every plan

Identity verification

Sign the user ID with HMAC-SHA256 on your server so a logged-in customer is provably who they say.

Growth+

CLI

List, update and reply to conversations from your terminal, or script them in CI.

Every plan

Hosted MCP server

Connect Claude, Cursor or any MCP client with OAuth 2.1. Install and set up Pengo on every plan; work the inbox, contacts and replies on Growth.

Growth+

AI connectors

Pengo Intelligence calls your API mid-chat with signed requests, so it can look things up and act.

02 Copy, paste, ship

Starting points that run as written

REST calls, signature checks, the widget API and the CLI. Swap in your key and go.

List open conversations with an API key from Settings → API keys.

curl "https://app.pengo.chat/api/v1/conversations?status=open&limit=20" \
  -H "Authorization: Bearer $PENGOCHAT_API_KEY"
03 Events and automation

Webhooks you can build a business on

Every delivery is signed and timestamped. Failures retry with exponential backoff, and anything that still fails lands in a dead-letter queue instead of vanishing.

Delivery log · chat.rated

Events

chat.createdchat.assignedchat.status_changedchat.resolvedchat.closedchat.ratedchat.ai_resolvedmessage.sentlead.createdlead.hotbooking.created

Workflows: trigger → conditions → actions

01 · When

Chat startedVisitor sends a messageIntelligence classifies the chatVisitor waits for a replySLA target missedLead becomes hotMeeting booked

02 · If

Intent is…Sentiment is…Urgency is…Message contains…Conversation field equals…Lead score ≥…

03 · Then

Wait up to 30 daysTag or assignSend saved replyDraft a reply with AISummarize to SlackSend to a webhook

Workflows run on every channel, so website chat, email and SMS follow the same rules.

04 REST API v1

The endpoints you’ll reach for first

Bearer-token auth, JSON in and out, pagination on every list. The full reference covers every request and response.

GET /v1/conversations List conversations; filter by status or assignee, paginate.
GET /v1/conversations/:id One conversation with its details.
PATCH /v1/conversations/:id Update status, assignee, priority or tags.
GET /v1/conversations/:id/messages Read the message history.
POST /v1/conversations/:id/messages Send a message as an agent or bot.
POST /v1/conversations/:id/notes Add an internal note for your team.
GET /v1/contacts List and search contacts; POST creates one, PATCH /v1/contacts/:id updates it.
GET /v1/visitors List visitors and their metadata; PATCH /v1/visitors/:id updates one.
GET /v1/agents Your teammates, for assigning and reporting.
GET /v1/tags Every tag in the workspace.
POST /v1/knowledge/documents Push knowledge for Pengo Intelligence; list sources and documents, or delete one.
GET /v1/audit-logs Admin activity, for your own security tooling.
05 AI-native

Built for the AI agents you already use

Give Claude, Cursor or any MCP client Pengo through a hosted MCP server, and give Pengo Intelligence your backend through signed connectors.

OAuth 2.1

Hosted MCP server

On every plan, your AI coding agent installs Pengo: it detects the platform, adds the widget, allows your domains and checks it’s live. On Growth, it works your inbox: lists and reads conversations, sends replies, adds notes and finds contacts.

{
  "mcpServers": {
    "pengo": {
      "url": "https://app.pengo.chat/api/mcp"
    }
  }
}
// Your client opens a Pengo sign-in (OAuth 2.1). Prefer a local server?
// npx -y @pengochat/mcp-server with PENGOCHAT_API_KEY set.
Growth+

Signed AI connectors

Register an HTTPS endpoint and Pengo Intelligence calls it mid-conversation. Requests are HMAC-signed, SSRF-checked and time-limited, and each call runs on its own or waits for your approval.

// Your endpoint, called by Pengo Intelligence mid-chat
app.post('/pengo/order-lookup', express.raw({ type: '*/*' }), (req, res) => {
  if (!verify(req.body, req.get('X-Pengo-Signature-V2'), SECRET)) {
    return res.sendStatus(401);
  }
  const { payload } = JSON.parse(req.body);
  res.json(lookupOrder(payload.email));
});
06 By plan

Start building on any plan

The widget API, identity verification and the MCP install tools come with Free. Webhooks and the full API join on Growth, and workflows on Scale.

Surface FreeStarterGrowthScaleEnterprise
REST API v1 –Not included –Not included Included Included Included
API keys –Not included –Not included Included Included Included
Signed, retried webhooks –Not included –Not included Included Included Included
Workflows –Not included –Not included –Not included Included Included
Widget JS API Included Included Included Included Included
Identity verification Included Included Included Included Included
CLI –Not included –Not included Included Included Included
Hosted MCP server Included Included Included Included Included
AI connectors –Not included –Not included Included Included Included

The hosted MCP server installs Pengo on every plan and works your inbox from Growth. Webhooks guide →

07 Questions

Developer FAQ

Which plans include the API?
The widget JavaScript API, identity verification and the MCP server’s install tools work on every plan. Webhooks, the REST API, API keys, the CLI and the MCP server’s inbox tools start on Growth, and workflows on Scale.
How do I verify a webhook?
Each delivery carries an X-Pengo-Signature-V2 header with a timestamp and an HMAC-SHA256 of timestamp.body using your webhook secret. Recompute it, compare in constant time, and reject old timestamps. A legacy X-Pengo-Signature header is also sent.
What happens if my endpoint is down?
Deliveries are retried with exponential backoff. If they keep failing they move to a dead-letter queue instead of disappearing, so you can see what was missed.
What can the MCP server do?
Pengo hosts it for you at app.pengo.chat/api/mcp. Connect Claude, Cursor, VS Code, Windsurf or any MCP client and sign in with OAuth 2.1, or use an API key. On every plan it installs Pengo for you: it detects your platform, adds the widget, sets allowed domains and checks the install. On Growth it also works your inbox: list and read conversations, send replies, add notes and find contacts.
How does identity verification work?
Your server computes HMAC-SHA256 of the user ID with your property’s identity secret and passes it as userHash to pengo('identify'). Verified users keep the same conversations on every device, and you can require verification so unverified calls are rejected.
Can Pengo Intelligence call our own API?
Yes, on Growth and above. Add a custom connector with your HTTPS endpoint and Pengo Intelligence can call it mid-conversation to look something up or make a change. Every request is HMAC-signed so your server can verify it, and you choose whether each call runs on its own or waits for a teammate’s approval. How custom connectors work.
Is there rate limiting?
Yes. The API and widget endpoints are rate limited, so every workspace stays fast and no one key can flood the platform.

Build on Pengo with a free account

Install in five minutes. Your AI agent is answering before your coffee cools.

Free plan forever · 14-day Growth trial · Cancel anytime

Start free