API
Use an organisation API key (Settings → API keys), sent as a bearer token. Limit: 60 requests
per minute per key. Errors look like { "error": { "code": "...", "message": "..." } }.
export SEENKITE_API_KEY=<your key> curl -H "Authorization: Bearer $SEENKITE_API_KEY" https://seenkite.com/api/v1/sites
Machine-readable spec: /api/v1/openapi.json
| Method | Path | What it does |
|---|---|---|
GET | /sites | List sites |
GET | /sites/{siteId} | Get a site |
POST | /sites/{siteId}/runs | Start an audit (counts against the monthly quota) |
GET | /runs/{runId} | Run status and step events |
GET | /runs/{runId}/report.json | Report (JSON) |
GET | /runs/{runId}/report.md | Report (Markdown) |
GET | /sites/{siteId}/reports/latest.md | Latest report (Markdown) |
GET | /sites/{siteId}/fixes | List fixes |
POST | /fixes/{fixId}/verify | Run the fix’s verify check against the live site |
POST | /fixes/{fixId}/applied | Mark a fix as applied |
GET | /sites/{siteId}/outcomes | Before/after metrics for applied fixes |
Typical agent loop
- Fetch the latest report:
GET /sites/:siteId/reports/latest.md - List open fixes to get each fix's
uuid:GET /sites/:siteId/fixes - Apply a fix in the code, deploy, then
POST /fixes/:uuid/verify - When it passes,
POST /fixes/:uuid/appliedwith the commit hash.
MCP server
AI agents that speak the Model Context Protocol (Claude Code, Cursor, VS Code, Codex and others) can use Seenkite directly, with the same API keys. Read-only keys only see the read tools. Exact setup for each agent is in Settings → API keys.
claude mcp add --transport http seenkite https://seenkite.com/mcp \ --header "Authorization: Bearer $SEENKITE_API_KEY"
| Tool | What it does | Key |
|---|---|---|
list_sites | Sites this key can access, with open and critical fix counts, AI crawler blocks and the last audit time. Start here. | Any |
get_site_overview | Where a site stands: fixes by severity, AI crawler access, Search Console and PageSpeed numbers from the latest audit, and the five fixes to do next. | Any |
list_fixes | Fixes for a site, highest priority first. Each has a fix_uuid for get_fix, verify_fix and mark_fix_applied. | Any |
get_fix | Everything needed to apply one fix: where it is, the current and proposed value, why it matters, evidence and how it is verified. | Any |
verify_fix | Checks the live site to see whether a fix is in place (e.g. the title now exists). Run it after deploying the change. | Read and write |
mark_fix_applied | Records that a fix was applied (optionally with the commit), and starts measuring its effect on Search Console clicks. | Read and write |
start_audit | Starts a new audit of a site (counts against the monthly limit). Returns a run_id for get_audit_status. Audits take 2–10 minutes. | Read and write |
get_audit_status | Status and step-by-step progress of an audit started with start_audit. | Any |
get_latest_report | The full latest audit report in Markdown: summary, AI crawler access, every fix with evidence, and content opportunities. | Any |
get_content_opportunities | Searches the site almost ranks for (positions 5–20 in Google), with the page to improve and a brief when available. | Any |
get_ai_visibility | Which AI crawlers can read the site, whether AI answers cite it for its queries, and how it compares with competitors. | Any |
get_outcomes | For each applied fix: the page's Google clicks for 28 days before and after, and a verdict adjusted for the site-wide trend. | Any |