Connect your agent

Authorize an AI agent to work in BitBoard. Connected agents read and edit your dashboards through the Model Context Protocol.

Your MCP URL

https://app.bitboard.work/mcp

You can also copy this from BitBoard under Settings → Agent access.

Connect your agent

Claude

1. Open Claude → Customize → Connectors.
2. Click +, then Add custom connector.
3. Paste your MCP URL.
4. Authorize BitBoard when prompted.

Claude MCP docs

Claude Code

Run this in your terminal:

claude mcp add --transport http bitboard https://app.bitboard.work/mcp

A browser window will open to authorize BitBoard.

Claude Code MCP docs

Cursor

One-click install

Set up manually

1. Open Cursor Settings (Cmd+Shift+J) → MCP → + Add New MCP Server.

2. This opens ~/.cursor/mcp.json. Add BitBoard to the mcpServers section:

   "bitboard": {
     "url": "https://app.bitboard.work/mcp"
   }

3. Save the file. A browser window will open to authorize BitBoard.

Cursor MCP docs

ChatGPT

1. Open ChatGPT → Settings → Apps.
2. Click Advanced settings and enable Developer mode.
3. Go back, then click Create app.
4. Enter name BitBoard, set Authentication to OAuth, and paste your MCP URL.
5. Click Create.

ChatGPT MCP docs

Codex

Codex stores MCP servers in ~/.codex/config.toml. Add BitBoard:

[mcp_servers.bitboard]
enabled = true
url = "https://app.bitboard.work/mcp"

Then authenticate:

codex mcp login bitboard

A browser window opens for OAuth consent. Codex stores the token and refreshes it on its own from then on.

To confirm BitBoard is wired up, run /mcp inside the Codex TUI and you should see bitboard with its tools loaded.

Codex MCP docs

BitBoard implements the Model Context Protocol over OAuth 2.0. Agents not listed here can still connect: paste the MCP URL into any MCP-compatible client.

Add data through your agent

Once your agent is connected, it can load data into a dashboard two ways. Every successful add becomes a named source in the dashboard’s catalog, queryable from any SQL block.

Push directly. When the data is already in the agent’s context (a small table from a previous step, a hand-curated CSV), dashboard_push_data writes it inline. Use dashboard_append_rows to extend an existing table without re-sending what’s there.

Upload a local file. Call dashboard_create_upload_url to mint a short-lived signed PUT URL. The agent’s host streams the file directly to that URL, then dashboard_finalize_upload converts it (typically to Parquet) and registers it as a source. Use this when the file lives on the agent’s machine and is too large to push inline.

For ongoing access to live data (your warehouse, your application database), see Connect your data.