MCP Quickstart
Connect Claude, ChatGPT, Cursor or Claude Code to your Specala AI meetings — paste one URL and sign in, or run the local npm server.
The Specala AI MCP server lets an AI assistant read your transcriptions directly. Once connected, you can ask things like "summarise my last call with Acme" or "what were the action items from yesterday's standup?" — the assistant calls Specala AI for you.
What is MCP?
The Model Context Protocol is an open standard that lets AI apps connect to external tools. Specala AI ships an official MCP server, so any MCP-compatible assistant can use your transcriptions.
There are two ways to connect. Most people want the first one.
| Remote connector (recommended) | Local server (npm) | |
|---|---|---|
| Setup | Paste one URL, sign in | Install via npx, paste an API key |
| Needs Node.js | No | Yes (18+) |
| Needs an API key | No — you sign in with your Specala AI account | Yes (sk_live_…) |
| Works in | claude.ai, ChatGPT, Claude Desktop, Claude Code, Cursor… | Claude Desktop, Claude Code, Cursor… |
| Saving files to your disk | No (content is returned in chat) | Yes — output_path + bulk ZIP export |
Both require a Specala AI plan with API & MCP access (Pro or Expert).
Option 1 — Remote connector (no install)
The hosted server lives at:
https://mcp.specala.ai/mcpYou add this URL to your assistant, it opens a Specala AI page in the browser, you sign in, pick a workspace and click Allow. That's the whole setup — no keys to copy, nothing to install.
Add the connector
- Open Settings → Connectors and click Add custom connector.
- Name:
Specala AI. Remote MCP server URL:https://mcp.specala.ai/mcp. Leave the advanced OAuth fields empty. - Click Add, then Connect — a Specala AI page opens in the browser.
MCP connectors are available on Plus/Pro/Team plans:
- Open Settings → Connectors and add a new connector.
- MCP server URL:
https://mcp.specala.ai/mcp, authentication: OAuth. - Save and connect — a Specala AI page opens in the browser.
One command:
claude mcp add --transport http specala-ai https://mcp.specala.ai/mcpThen run /mcp inside Claude Code and pick specala-ai → Authenticate to sign in.
Add the server to your MCP config:
{
"mcpServers": {
"specala-ai": {
"url": "https://mcp.specala.ai/mcp"
}
}
}Cursor will prompt you to authenticate in the browser on first use.
Sign in and allow access
On the Specala AI page that opens: sign in (if you aren't already), pick the workspace the assistant may read, and click Allow.
The assistant gets read-only access to that one workspace — it can never edit or delete anything. You can disconnect at any time from the assistant's settings, and the access token expires and renews automatically in the background.
Ask your assistant
Start a new chat and try:
Use Specala AI to find my meetings from this week and summarise the action items.
The assistant will call specala_list_transcriptions, then specala_get_transcription, and answer
using your real transcripts.
Option 2 — Local server (npm)
Choose this when you want the assistant to save files to your disk (per-meeting .md/.txt
exports, bulk ZIP export of up to 100 meetings) or when you prefer key-based auth for scripts and
CI. Requires Node.js 18+ and an API key.
Create an API key
In the Specala AI web app, open Settings → API & MCP and click Create key. Copy it
immediately — it starts with sk_live_ and is shown only once.
See Authentication for key scope and security details.
Create API keyAdd the server to your assistant
Open the config via Claude → Settings → Developer → Edit Config and paste:
{
"mcpServers": {
"specala-ai": {
"command": "npx",
"args": ["-y", "specala-mcp-server"],
"env": {
"SPECALA_API_KEY": "sk_live_your_key_here"
}
}
}
}Restart Claude Desktop.
Add via Cursor Settings → MCP → Add new server, or paste into the config:
{
"mcpServers": {
"specala-ai": {
"command": "npx",
"args": ["-y", "specala-mcp-server"],
"env": {
"SPECALA_API_KEY": "sk_live_your_key_here"
}
}
}
}One command — no file editing:
claude mcp add specala-ai \
--env SPECALA_API_KEY=sk_live_your_key_here \
-- npx -y specala-mcp-serverLocal server configuration
The local server reads two environment variables:
| Variable | Required | Default | Description |
|---|---|---|---|
SPECALA_API_KEY | Yes | — | Your sk_live_ API key |
SPECALA_API_URL | No | https://app.specala.ai | Override only for self-hosted/staging |
What the assistant can do
Once connected, the assistant has these tools available:
| Tool | What it does | Remote | Local |
|---|---|---|---|
specala_list_projects | Discover projects (for filtering) | ✓ | ✓ |
specala_list_transcriptions | Find meetings — filter by date, project, language | ✓ | ✓ |
specala_get_transcription | Read full text + AI reports for one meeting | ✓ | ✓ |
specala_export_transcription | Get one meeting as .md / .txt | ✓ (in chat) | ✓ (saves to disk) |
specala_bulk_export | Export up to 100 meetings as files in one call | — | ✓ |
See the MCP Reference for every parameter.