Speak AI Docs
Connect an AI agent, call the REST API, or install the SDK. Written for people, and built so AI assistants can read every page too.
4.9 on G2Trusted by 250,000+ teams
Start building
Which surface are you building against?
Speak AI is one system of record for voice, video, and text data. Pick the surface you work through, and start with the first command you run.
Connect an agent to your workspace
The Speak AI MCP server gives Claude, ChatGPT, Cursor, and VS Code 112 tools covering media, transcripts, AI insights, folders, recorders, automations, and exports. Most clients need one URL and one approval.
https://api.speakai.co/v1/mcp
claude mcp add speakai --transport http --url https://api.speakai.co/v1/mcpCall the REST API from your own stack
The Speak AI API covers 70 endpoints across 13 resources for uploads, transcripts, insights, folders, recorders, webhooks, and automations. Exchange your API key for an access token, then send that token on every call.
https://api.speakai.co/v1
curl -X POST 'https://api.speakai.co/v1/auth/accessToken' \
-H 'x-speakai-key: sk_test_speak_0000000000000000' \
-H 'Content-Type: application/json' \
-d '{}'Use the package as a Node library
The package that runs the MCP server is also a Node library. Install it to register Speak AI's tools, resources, and prompts on your own MCP server, or to call the API through an authenticated client. It needs Node 22 or newer.
@speakai/mcp-server
npm install @speakai/mcp-serverTrack what shipped, in a feed
The changelog records every feature, fix, and improvement Speak AI has shipped, grouped by year. Point a feed reader at the RSS file, or pull the JSON feed and render the entries inside your own app.
/changelog/feed.json
curl https://docs.speakai.co/changelog/feed.jsonWhat else is in these docs?
Three more sections cover the questions the reference pages do not: how the product behaves, what each tool does, and what changed recently.
Reading this as an agent?
Every page on this site is published as clean Markdown alongside the HTML. Fetch llms.txt for the index of every page, or llms-full.txt for the whole corpus in one document. Append index.md to any page path to read that page on its own.