NDASentry runs as an MCP server. Add it to Claude Desktop, Cursor, Continue, or Cline, and any NDA you send to your AI assistant can be scored against the ten-category risk taxonomy — without leaving the conversation.
Each AI client stores MCP server configurations in its own JSON file. Click Copy config for your client, then jump to Step 2 to paste it in.
Don't see your client? NDASentry follows the MCP standard and works with any compliant client. See the MCP clients list.
mcp-remote proxy because the
stable release uses stdio transport. Requires Node.js installed.
experimental.modelContextProtocolServers block.
~/.config/cline/.
Use the File path button on your client's card above to see the exact location. If the file doesn't exist yet, create it.
If you already have other MCP servers configured, copy just the
"ndasentry" entry into your existing "mcpServers" object.
Otherwise paste the whole snippet.
Make sure the file remains valid JSON — check that braces and commas line up. Most editors will highlight syntax errors.
Fully quit and reopen. Some clients reload MCP servers automatically; restarting always works.
After restarting your AI client, start a new conversation and try
a prompt like this. If NDASentry is connected, your AI will call
the preview_nda_risk tool and return a structured
analysis.
Drop your PDF on ndasentry.ai for a one-time, account-free analysis. Same scoring engine, $9, no install required. The document evaporates after the report loads.