Claude.ai (web)
- Open claude.ai/settings/connectors
- Click Add custom connector
- Paste the URL above
- Click Add, sign in to Speak AI, click Allow on the permission popup
No Terminal. No npm. No JSON config.
https://api.speakai.co/v1/mcp
Install
One-click connect is the easiest install (recommended for coaches and non-developers). Or paste an API key as a header for any client that prefers manual setup.
Paste the connector URL into your AI tool's connector settings — same idea as pasting a Zoom link into your calendar. For the manual API-key route, generate a key at app.speakai.co/developers/apikeys and add it as an Authorization header.
Expand Custom headers in the Add Connector dialog and add Authorization = Bearer <your-key>. Generate a key at app.speakai.co/developers/apikeys.
{
"type": "mcp",
"server_url": "https://api.speakai.co/v1/mcp",
"authorization": "Bearer YOUR_KEY"
}
claude mcp add --transport http speakai https://api.speakai.co/v1/mcp --header "Authorization: Bearer $SPEAKAI_KEY"
What's inside
Everything Speak AI can do — exposed for your AI assistant to call. You don't memorize them; the AI picks the right one.
Upload audio or video, fetch transcripts, get AI insights, organize, favorite, export.
Run AI prompts across one file, a folder, or your whole workspace. Save favorites.
Full-text search across transcripts, insights, and metadata. Workspace-level stats.
Create folders, save filtered views, bulk move files, share read-only collections.
Public recording links for clients, prompted question sets, branded submission pages.
Highlight clips from any time range. Export PDF, DOCX, SRT, VTT, TXT, CSV.
Schedule Speak AI to join a Zoom / Google Meet / Teams call and transcribe automatically.
Trigger actions on new recordings. Send events to your own backend.
Standalone AI-analyzed text notes plus custom fields you can tag any media file with.
What you can ask
Claude and ChatGPT pick the right ones based on what you ask. Just talk to them like a teammate.
Need a hand?
If anything feels confusing or doesn't work, we'll get you set up. Most coaches click the one-click connect button and approve the permission popup in under a minute. You don't need Terminal, you don't need to install anything on your computer, and you definitely don't need to hire a Virtual Assistant.
Works with
FAQ
Privacy, permissions, regions, and the developer path.
Either method grants the AI assistant read AND write access to your full Speak AI workspace — including the ability to view, modify, and delete recordings, folders, and clips. Claude/ChatGPT only access specific data when your AI assistant calls a tool to answer a question you asked. Your recordings stay in your Speak AI workspace; they are not copied or stored by Anthropic or OpenAI.
Yes — write access includes delete operations. We recommend always asking your AI assistant to confirm before deleting. MCP (the standard way Claude and ChatGPT connect to apps) marks delete tools as destructive, and most clients (including Claude) prompt you before invoking them.
This depends on your specific situation, your existing Business Associate Agreements, and the contents of your recordings. When you grant the connector access, transcript content (including participant names and clinical detail) gets sent to Anthropic / OpenAI as part of the AI's reasoning. If your work is HIPAA-regulated, please review your participant consent practices before installing, and check whether your existing BAAs cover this use. Email privacy@speakai.co with specific compliance questions.
Three options, all live today: (1) remove the connector inside Claude/ChatGPT itself, (2) for one-click connect installs, revoke the connection at api.speakai.co/v1/oauth/connections, or (3) for API-key installs, rotate or revoke the key at app.speakai.co/developers/apikeys — any tool using that key loses access immediately.
Yes for Claude (.ai web, Desktop, Code) and Cursor / VS Code — all of those work everywhere. ChatGPT custom connectors are not available in those regions (an OpenAI restriction, not ours). If you want ChatGPT specifically, you'll need to use the API / Responses path with an API key; otherwise we recommend Claude.ai or Claude Desktop, which install with one click.
Speak AI's built-in Magic Prompt is great for asking questions inside Speak. This connector lets you ask the same kinds of questions inside Claude or ChatGPT, where you might already be doing other work — drafting emails, planning content, debugging code. Use whichever feels more natural for the task.
No new app required — Speak AI hosts the connector; Claude/ChatGPT contact it directly. The one-click connect install runs entirely in your browser (paste a URL, click Allow on the permission popup). The manual API-key alternative is also paste-only — one URL plus one Authorization: Bearer <your-key> header, no local install. The npm package (@speakai/mcp-server) exists for developers who want to self-host or use a CLI — that's not the path for the average customer.
See the README on GitHub — the developer reference is collapsed at the bottom and includes the CLI commands, environment variables, REST API auth flow, and stdio-mode setup for clients that don't speak remote HTTP.
Single-attempt failures are usually transient — try the same prompt again. If it persists, the connector may need to be reinstalled: for one-click connect installs, revoke at api.speakai.co/v1/oauth/connections then reinstall; for API-key installs, regenerate your key at app.speakai.co/developers/apikeys and update the connector. If still broken, email accounts@speakai.co.
Some Claude Team / Enterprise workspaces restrict custom connectors via admin policies. Ask your workspace admin to allow https://api.speakai.co/v1/mcp in Organization Settings → Connectors.
Privacy & data
The short version of how data flows when you connect.
You're authorizing that AI assistant to read and modify your Speak AI workspace on your behalf — including media files, transcripts, and AI insights.
For data-handling questions, see speakai.co/privacy or email privacy@speakai.co.