Run Locan from Claude, OpenClaw, Cursor or n8n

The same local SEO tools behind locan.ai, callable from an AI agent or a workflow tool over MCP, a CLI, or plain HTTP.

What it is

An MCP server and a CLI over the Locan API

The locan-tools package installs two thin HTTP clients: locan, a CLI, and locan-mcp, an MCP server exposing the same eight tools over stdio (or streamable HTTP with locan-mcp --http --port 8765). Neither holds a Google Places or LLM key — both call the Locan API, hosted or self-hosted.

Claude Code

Add the MCP server with one command

claude mcp add locan -- uvx --from locan-tools locan-mcp

Claude Desktop / Cursor

Or configure it as a JSON MCP server

{
  "mcpServers": {
    "locan": {
      "command": "uvx",
      "args": ["--from", "locan-tools", "locan-mcp"],
      "env": { "LOCAN_API_URL": "https://api.locan.ai/api" }
    }
  }
}

Point LOCAN_API_URL at a self-hosted instance to skip the hosted daily limits — see self-hosting Locan.

Reference

The eight tools

MCP tool	What it returns	Needs
find_business	Candidate matches: placeId, name, address	Places key
gbp_audit	Score /100, grade, per-factor checks, AI suggestions	Places key + LLM
gbp_categories	Category comparison and add/keep/review recommendations	Places key
rank_grid	Grid of map ranks, summary, outranking competitors	Places key
ai_visibility	AI mention score, per-engine results, missing topics	LLM
review_reply	Three draft review replies plus handling tips	LLM
review_link	Direct Google review link for a place id	None
schema_jsonld	LocalBusiness JSON-LD built from the Google profile	Places key

The CLI exposes the same eight operations as hyphenated commands (find-business, gbp-audit, gbp-categories, rank-grid, ai-visibility, review-reply, review-link, schema-jsonld) — see cli/README.md .

For agent workflows

Building an audit skill

skills/locan-local-seo/SKILL.md in the repository is a ready-made procedure for auditing a business with these tools: resolve the business, run the audit, check categories and rankings, check AI visibility, then write a prioritised fix list. Drop it into an agent's skills directory, or read it as a reference for the call order and defaults.

Prefer a workflow tool instead of an agent? See running the profile optimizer from n8n.

View the skill on GitHub

Questions about the MCP server and CLI

What does a rank_grid call cost? +

One Google Places request per grid cell: 9 for the default 3×3 grid, 25 for a 5×5 grid. It is the most expensive of the eight tools.

Does any of this change my Google Business Profile? +

No. Every tool reads public Google data; none of them can edit a profile. Every recommendation is applied by you, in your own Google Business Profile or website.

Can I self-host to remove the daily limits? +

Yes. Point LOCAN_API_URL at your own instance (for example http://localhost:8080/api) and the hosted daily limits no longer apply. See the self-host guide.

Which AI models does this use? +

The hosted API uses a Replicate-hosted model. A self-hosted instance can use Replicate, OpenAI directly, or any OpenAI-compatible endpoint such as Ollama via OPENAI_BASE_URL.