CitationOne MCP server
Run GEO audits and read their results in a conversation with an AI assistant - in Claude, Cursor and other MCP clients. The same engine, API key and credits as in the dashboard.
Sample conversation
Audit https://example.com/blog/geo for the keyword "what is geo" and list the three fixes with the biggest impact.
Audit started. It takes a few minutes - checking progress.
Done: Content Quality Score 61/100. Biggest wins: answer in the first sentence of the definition section, a GEO vs SEO table and sources next to the study data. Full report: app.citationone.com/…
Connect in a minute
You need an API key (in the dashboard: Settings → API) and the server URL. A key allowed to submit audits lets the assistant start them; a read-only key is enough to browse results.
claude mcp add --transport http citationone https://app.citationone.com/api/mcp \
--header "Authorization: Bearer co_live_..."{
"mcpServers": {
"citationone": {
"url": "https://app.citationone.com/api/mcp",
"headers": { "Authorization": "Bearer co_live_..." }
}
}
}{
"mcpServers": {
"citationone": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://app.citationone.com/api/mcp",
"--header", "Authorization:${AUTH_HEADER}"],
"env": { "AUTH_HEADER": "Bearer co_live_..." }
}
}
}Tools available to the assistant
| Tool | What it does | API equivalent |
|---|---|---|
start_audit | Audits a URL or pasted text (e.g. a draft before publishing), optionally with a keyword, language and project. Returns a job id. | POST /api/v1/audits |
get_audit | Status while running; when finished, a summary: CQS against the top 10 average, dimensions, the 10 key recommendations and the report link. Full result with before/after text on request. | GET /api/v1/audits/{id} |
list_audits | Recent audits with scores, filtered by project, domain or status. | GET /api/v1/audits |
get_credits | Credit balance and recent credit events. | GET /api/v1/credits/usage |
An audit usually takes 3-8 minutes. The assistant starts it and then checks progress every 30-60 seconds on its own.
How it works
The same key as the API
The MCP server uses the API keys from your dashboard. Key scopes work the same way, and revoking a key cuts off the assistant too.
1 audit = 1 credit
An audit started by the assistant costs the same as in the dashboard and over the API, from the same credit pool. Reading results, the audit list and the balance is free. The first three audits on a new account are free.
A summary for the chat, full result on request
By default the assistant gets a compact summary that fits in a conversation. It fetches the full result with before/after text when you ask for it.
A link to the full report
Every result includes the report URL in the dashboard - with charts, the competitor comparison and PDF export.
The same limits and safeguards
Rate limits, concurrent audits and duplicate-submission protection work as in the API. Errors reach the assistant as one sentence with a code.
Page text treated as data
Results quote text from the audited page and competing pages. The server tells the assistant to treat that text as material to analyse.
Claude skill - audits in agency work
The MCP server gives the assistant the tools; the CitationOne skill teaches it how to use them in client work. It works in Claude Code, Claude Desktop and claude.ai; in Cursor the MCP server alone is enough.
- confirms the cost and credit balance before every audit,
- checks whether the page was audited in the last few days before spending a credit,
- reads the score against the Google top 10 average and treats gaps under 5 points as level,
- merges audits of several pages of one client into one priority list with site-wide fixes,
- compares audits before and after changes,
- writes a client summary from a ready template.
mkdir -p ~/.claude/skills/citationone
curl -o ~/.claude/skills/citationone/SKILL.md \
https://app.citationone.com/skills/citationone/SKILL.mdIn claude.ai, zip the citationone folder with this file and upload it in the skills settings.
Download the skill (SKILL.md) ↓You create API keys in the dashboard: Settings → API. The documentation has the details and the full parameter list.