MCP Server (AI Integration)

The Donotname MCP (Model Context Protocol) server lets you connect AI tools like Claude Desktop, Claude Code, or any MCP-compatible client directly to your base. The AI can then read, search, create, and update your records conversationally.

What is MCP?

MCP (Model Context Protocol) is an open standard that lets AI assistants securely connect to external data sources. Instead of copy-pasting data into a chat, the AI can directly access your tables, search records, and traverse your knowledge graph.

Setup

1. Create a Personal Access Token

Go to Settings > Access Tokens
Click Create Token
Name it (e.g., "MCP Server")
Copy the token — it is shown only once

2. Configure Claude Desktop

Add the following to your Claude Desktop MCP configuration file (claude_desktop_config.json):

{
  "mcpServers": {
    "donotname": {
      "command": "npx",
      "args": ["@donotname/mcp-server"],
      "env": {
        "DONOTNAME_API_URL": "https://app.donotname.com",
        "DONOTNAME_API_TOKEN": "your-personal-access-token"
      }
    }
  }
}

3. Configure Claude Code

Run the MCP server from your terminal:

DONOTNAME_API_URL=https://app.donotname.com \
DONOTNAME_API_TOKEN=your-token \
npx @donotname/mcp-server

Or add it to your .claude/settings.json for automatic startup.

Available Tools

Once connected, the AI has access to these tools:

Read-Only Tools

Tool	Description
`list_bases`	List all bases you have access to
`list_tables`	List all tables in a base
`get_table_schema`	Get field definitions for a table
`list_records`	List records with pagination
`get_record`	Get a single record with all fields
`search_records`	Search records by title across all tables

Knowledge Graph Tools

Tool	Description
`get_backlinks`	Find records that link to a specific record via wiki links
`get_knowledge_context`	Traverse wiki-link graph (1-3 hops) to find related records
`get_knowledge_graph`	Get the full graph structure (nodes + edges) for visualization

Write Tools

Tool	Description
`create_record`	Create a new record in a table
`update_record`	Update fields on an existing record
`delete_record`	Delete a record

Security

User-level permissions only — the MCP server authenticates with your personal access token and gets exactly your permissions. It cannot access data you cannot access.
No elevated privileges — the server runs as a separate process and calls the same REST API that you use in the browser.
Token-based — revoke the token at any time in Settings to disconnect the MCP server.
No performance impact — the MCP server runs outside the main application, so it does not affect other users.

Example Conversations

Searching your data:

"Find all records mentioning Acme Corp" AI uses search_records to find matching records across all tables

Understanding context:

"What is connected to the Q4 Proposal?" AI uses get_knowledge_context to traverse wiki links and return related meeting notes, contacts, and deals

Creating records:

"Add a new task: Follow up with Acme Corp by Friday" AI uses create_record to insert a new record in your Tasks table

Updating data:

"Mark the Acme Corp deal as Closed Won" AI uses search_records to find the record, then update_record to change the Status field

Environment Variables

Variable	Required	Description
`DONOTNAME_API_URL`	Yes	Your Donotname instance URL (e.g., `https://app.donotname.com`)
`DONOTNAME_API_TOKEN`	Yes	Personal access token from Settings > Access Tokens

Troubleshooting

"DONOTNAME_API_TOKEN is required" — set the environment variable before running the server
401 Unauthorized — your token may be expired or revoked. Create a new one in Settings.
403 Forbidden — you do not have permission to access the requested resource. Check your role in the space/base.
Connection refused — verify the DONOTNAME_API_URL is correct and the server is running.