Build Siesta AI into anything

Wherever your products, scripts, and other AI tools live, Siesta AI meets them there. Reach its agents, chat, and knowledge through a REST API, an embeddable web widget, a command-line tool, an MCP server, an agent-to-agent protocol, and a realtime streaming API.

REST API Web Widget CLI MCP A2A Realtime API

Read the docs API reference

Terminal

# Ask an agent, straight from your stack
$ curl https://api.siesta.ai/v1/chat \
  -H "Authorization: Bearer $SIESTA_KEY" \
  -d '{
    "agent_id": "sales-assistant",
    "message": "Summarize the new leads from today"
  }'

# → streams a grounded answer from your data

Build on Siesta AI, your way

Whether you’re shipping a product feature, wiring up an internal script, or connecting another AI system, there’s a clean interface waiting for it.

REST API

Chat, agents, webhooks, and data sources over a simple, predictable HTTP API. The foundation everything else is built on.

Web Widget

Drop a fully-functional AI chat onto any website or portal with a single script tag. No build step required.

CLI

Run agents, manage knowledge, and script Siesta AI straight from your terminal, pipelines, and cron jobs.

MCP Server

Expose your workspace over the Model Context Protocol so Claude, Cursor, and any MCP client can use your agents and data as tools.

Agent-to-Agent

Let external agents discover and delegate work to your Siesta AI agents over the open A2A protocol.

Realtime API

Open a live agent session and stream audio, transcripts, and tool events over WebSocket for voice and realtime interfaces.

REST API

The API everything runs on

Send messages, spin up and configure agents, react to events with webhooks, and manage the knowledge your agents can reach, all through one clean, well-documented REST API that any language can call.

Chat & conversations Agents API Webhooks Data sources

chat.js

// Send a message to one of your agents
const res = await fetch("https://api.siesta.ai/v1/chat", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.SIESTA_KEY}`,
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    agent_id: "support-bot",
    message: "How do I reset my password?"
  })
});

const { answer } = await res.json();

Web Widget

Add AI to any site in one line

Paste a single script tag and a Siesta AI chat bubble appears on your public site, help center, or client portal, answering from your content, matching your brand, and handing off to your agents when it matters.

One-line embed On-brand styling Grounded in your content No support ticket needed

index.html

<!-- Drop this before </body> -->
<script
  src="https://widget.siesta.ai/v1.js"
  data-agent="support-bot"
  data-theme="brand"
  defer></script>

<!-- Done. The chat bubble is now live. -->

CLI

Siesta AI in your terminal

Install once and drive your whole workspace from the command line. Run agents, upload knowledge, and pipe results into scripts. Perfect for automation, CI pipelines, and scheduled jobs.

Run agents Manage knowledge Scriptable output CI & cron friendly

Terminal

# Install and sign in
$ npm i -g @siesta-ai/cli
$ siesta login

# Run an agent and capture the result
$ siesta agents run sales-assistant \
    --input "Draft a follow-up for Acme Corp" \
    --json > reply.json

# Add a document to your knowledge base
$ siesta knowledge add ./handbook.pdf

MCP Server

Plug your workspace into any AI tool

Siesta AI speaks the Model Context Protocol. Point Claude, Cursor, or any MCP-compatible client at your workspace and your agents, knowledge, and tasks instantly become tools they can call, with no glue code to maintain.

Works with Claude & Cursor Agents as tools Knowledge search Zero glue code

mcp.json

// Add Siesta AI to your MCP client
{
  "mcpServers": {
    "siesta": {
      "url": "https://mcp.siesta.ai",
      "headers": {
        "Authorization": "Bearer ${SIESTA_KEY}"
      }
    }
  }
}

Agent-to-Agent

Your agents, in the wider network

Every Siesta AI agent can publish an open agent-to-agent (A2A) profile, so other systems can discover its skills and delegate real work to it, and your agents can call out to external ones the same way. Multi-agent workflows without a proprietary lock-in.

Open protocol Skill discovery Delegate & receive tasks Multi-agent workflows

agent-card.json

# Any agent exposes a discoverable profile
GET /.well-known/agent.json

{
  "name": "Siesta Sales Agent",
  "protocol": "a2a/v1",
  "skills": ["research", "draft", "schedule"],
  "endpoint": "https://a2a.siesta.ai/agents/sales"
}

Realtime API

Live agents, streamed as they think

Open a short-lived agent session and stream audio, transcripts, tool calls, and approval states over a single WebSocket. Built for voice experiences, live agent interfaces, and public widgets that need to react the moment things happen.

Streaming audio Live transcripts Tool & approval events WebSocket transport

realtime.js

// 1. Create a short-lived realtime session
const res = await fetch(
  `${API}/Agent/${agentId}/realtime-session`,
  { method: "POST", headers: { "X-Api-Key": KEY } }
);
const { sessionId, conversationId } = await res.json();

// 2. Stream transcripts, audio, and tool events
const ws = new WebSocket(
  `${WS}/Agent/${agentId}/realtime` +
  `?sessionId=${sessionId}`
);
ws.onmessage = (e) => render(JSON.parse(e.data));

Start building today

Grab an API key, read the docs, and ship your first Siesta AI integration in an afternoon.

Read the docs API reference