MCP server
Citevera is a Model Context Protocol server. Any MCP-capable assistant can run scans, quick checks and monitoring queries against it, with no code to write.
MCP is the open standard that lets assistants such as Claude, ChatGPT, Cursor and Claude Code call outside tools. Citevera publishes one endpoint:
https://api.citevera.com/mcp
Connecting asks you to sign in to Citevera and approve the assistant, the standard OAuth flow every MCP directory expects. No key to paste. A free account gets the free tier: quick checks, a report preview, and the findings list. A paid plan runs scans on your quota and returns full reports and fixes.
Two alternatives: an API key from app.citevera.com/settings/api-keys sent as a bearer token, for tools that cannot open a browser; and https://api.citevera.com/mcp/public, which needs no account at all and serves the free tier.
Tools
| Tool | What it does | Access |
|---|---|---|
| scan_site | Starts a full audit of a site and returns a scan id. | Free (rate limited) or plan quota |
| get_scan_status | Progress of a scan. | Free |
| get_report | Scores, funnel, priority plan and findings. Can wait up to 45 seconds per call for the scan to finish. | Preview free; findings with an email; full report on a paid plan |
| get_fix | One finding in full, including the fix body. | Paid plan |
| check_llms_txt | Validates llms.txt and llms-full.txt. | Free |
| check_ai_crawler_access | robots.txt and reachability per AI crawler. | Free |
| check_ai_brand_mentions | Whether AI assistants mention the site for its category. | Free, one result per domain per day |
| check_answerability | How directly a page answers likely questions. | Free |
| check_entity | Whether the organisation is stated clearly for AI models. | Free |
| monitoring_summary | Weekly citation rate and top cited URLs for tracked prompts. | Monitoring subscription |
| whoami | What this connection can do. | Free |
A typical session: ask the assistant to audit a site. It calls scan_site, polls get_report (45 seconds per call, most MCP clients time out at 60), then pulls get_fix for the items it wants to apply. Scans take one to three minutes.
Claude Code
claude mcp add --transport http citevera https://api.citevera.com/mcp
Then run /mcp inside Claude Code and choose Citevera to sign in; a browser window opens for the approval. To skip the browser, add --header "Authorization: Bearer cv_live_YOUR_KEY" instead.
Claude (claude.ai and desktop)
Settings, Connectors, Add custom connector, then paste the endpoint URL. Claude registers itself, sends you to Citevera to sign in and approve, and is connected on your plan.
Cursor, Windsurf, VS Code and other editors
Add to the editor's MCP configuration (for Cursor, ~/.cursor/mcp.json):
{
"mcpServers": {
"citevera": { "url": "https://api.citevera.com/mcp" }
}
}Editors that support OAuth open the sign-in page on first use. For one that does not, add "headers": { "Authorization": "Bearer cv_live_YOUR_KEY" } or point it at https://api.citevera.com/mcp/public.
ChatGPT
With developer mode enabled, Settings, Connectors, Create, then paste the endpoint URL and choose OAuth. ChatGPT sends you to Citevera to sign in and approve.
Limits and privacy
- Free-tier scans share the public limits: a few scans per day per address and a daily budget across all free scans. Paid keys use the plan's monthly audit quota.
- Quick checks share the website's tool budget of ten checks per hour per address.
- Approvals use short-lived access tokens with refresh, and a key can be revoked at any time from the settings page. Revoked credentials stop working immediately.
- Scan results are the same reports you see on scan.citevera.com; nothing extra is collected from the assistant.
Prefer a file-based workflow? The SKILL.md export still works for applying a report offline.