Connecting an AI agent

GeckoGrid ships with everything an AI agent needs to run your scanning workflow. Setup is one copy-paste, and it all stays on your machine.

What MCP is

Model Context Protocol (MCP) is an open standard that lets AI assistants call external tools in a controlled way. An app that runs an MCP server publishes its abilities as named tools, and a compatible agent can call those tools when you ask it to do something. In practice: you tell your agent “scan the Monday roster and summarize what moved”, and the agent reads your projects, queues the scans, and reports back.

Set it up

1. Open the AI & MCP page in GeckoGrid.
2. Click Generate token. The token is a 64-character key stored locally, and every agent request must carry it.
3. Copy the setup prompt the page generates. It contains the connection facts for your machine, endpoint and token included.
4. Paste the prompt into your agent, for example Claude, Cursor, or Codex, and restart that agent session.

The agent connects, lists the GeckoGrid tools, and loads its skills. No config-file editing is needed. One caveat: the setup prompt embeds your token, so if you share a conversation transcript, rotate the token on the AI & MCP page first.

What the agent gets

• 47 tools covering projects, scans, results, competitors, schedules, settings, exports, and reports. An agent can start and cancel scans, create schedules, geocode addresses, export CSVs and heatmap images, and build client and prospect reports in your branding.
• 10 analyst skills that teach interpretation rather than operation: rank diagnostics, competitor analysis, trend analysis, keyword research, GBP audit, website audit, schedule planning, prospect reports, prospect pipelines, and client reports.

Same machine, no cloud hop

The MCP server binds to localhost only, so it is unreachable from your network or other machines, and every call requires the bearer token. Your agent and the GeckoGrid app must run on the same computer. Projects and scan data live in GeckoGrid’s local database and never leave your machine through MCP.

If the connection fails, check that GeckoGrid is running (the server starts shortly after launch and stops when the app quits) and that the token in your agent config matches the AI & MCP page. Port collisions are handled automatically, and the page shows the actual endpoint in use.

Go deeper

• MCP tool reference: all 47 tools grouped by category
• MCP analyst skills: the ten interpretation skills and when an agent loads each one
• MCP recipes: paste-ready workflows for client reports, prospect pitches, rank checks, and competitor watch
• MCP security: the connection model in plain terms, localhost binding, token auth, and what an agent can never touch