MCP Server
Drive FieldTrip from your AI assistant.
FieldTrip's Model Context Protocol server is live. Connect it to Claude Desktop, Claude Code, Cursor, or ChatGPT as a remote connector. Workspace-scoped. Audit-logged. Signoff-required for any action that ships.
https://mcp.draftline.pro/mcpWhat you can do.
The MCP server publishes tool discovery at the endpoint — your client picks up the available tools on connect. The shape is the same shape your team uses inside the platform.
- Query jobs, routes, nodes, and connections
- Inspect classifications and confidence scores
- Trigger batch classification runs
- Pull the review queue and sign off on items
- Reassociate photos to corrected nodes
- Push writeback to your GIS (on a signoff)
Setup: Claude Desktop
- 1. Open Claude Desktop → Settings → Connectors.
- 2. Click Add custom connector.
- 3. Paste the FieldTrip MCP URL:
https://mcp.draftline.pro/mcp- 4. Authenticate when prompted — FieldTrip will mint a workspace-scoped token tied to your user.
Setup: Claude Code
From your terminal:
claude mcp add --transport http draftline https://mcp.draftline.pro/mcpOr add it to your project's .mcp.json:
{
"mcpServers": {
"draftline": {
"type": "http",
"url": "https://mcp.draftline.pro/mcp"
}
}
}Setup: Cursor
Add FieldTrip to ~/.cursor/mcp.json:
{
"mcpServers": {
"draftline": {
"url": "https://mcp.draftline.pro/mcp"
}
}
}Restart Cursor. FieldTrip tools show up in the agent picker.
Setup: ChatGPT
Use OpenAI's Connectors UI (rolling out broadly) or a Custom GPT pointed at the MCP URL.
- 1. ChatGPT → Settings → Connectors → Add custom connector.
- 2. Paste
https://mcp.draftline.pro/mcp. - 3. Authenticate when prompted.
If your ChatGPT account doesn't yet have native MCP connector support, ping sales@gofield.net — we'll get you on a Custom GPT workaround in the meantime.
Auth and scopes
The MCP server authenticates via bearer token. Tokens are workspace-scoped and carry the same permissions the issuing user has inside FieldTrip. An agent connecting on your token can't do anything you can't do.
Generate a token in your workspace settings under Developer → MCP tokens. Tokens can be scoped to a subset of permissions (read-only, classify-only, etc.) and rotated at any time.
Audit log on every action
Every MCP call shows up in your workspace audit log: which client made the call (Claude Desktop, Cursor, ChatGPT, custom), which token authorized it, what action ran, and the result. Same audit surface your QC leads already use for human actions.
What people use this for
- Status questions in plain English. "What's blocking the Buckhead Loop route?" → Claude pulls stalled stages, surfaces conflicts, suggests the next move.
- Triage and reassignment. Ask the agent to reassign reviews to available capacity, trigger reclassification on a slice of poles, or flag photos that look off.
- Programmatic exports. Pull classifications + audit log for a job into wherever your team works.
- Custom internal agents. Wire FieldTrip into your own agent framework — the MCP surface is the same one our purpose-built agent uses.
Stuck on setup or need a different transport?
Ping us. We'll get you connected — and if your client isn't in the list above, tell us which one and we'll document it.