Every workspace resource, over an API — or an agent
Projects, templates, sharing links, and audits, all reachable with the same workspace API key you already have. Script it with the REST API, or hand it directly to an AI agent over MCP.
A conventional JSON API, scoped to your workspace
/api/v1/projects, /api/v1/templates, /api/v1/links, and /api/v1/audits cover create, read, update, and delete for everything the dashboard can do — authenticated with a single Bearer token.
$ curl -X POST https://useopengraph.com/api/v1/audits \
-H "Authorization: Bearer og_live_..." \
-d '{"projectId":"prj_8k2","domain":"acme.dev"}'
{
"auditId": "aud_4f01",
"status": "queued",
"pollUrl": "/api/v1/audits/aud_4f01"
}
- Full CRUD, not a read-only export — create templates, publish versions, manage links from your own code
- Ownership-checked on every request — a key only ever sees its own workspace's data
- Async by design where it matters: triggering an audit returns instantly, poll for status
Hand your workspace to an agent, not just a script
The same operations as the REST API, exposed as MCP tools — list_projects, create_template, trigger_audit, get_audit, and more — so Claude or any MCP-compatible agent can act on a workspace directly, with the same auth and the same usage metering.
→ trigger_audit({ projectId: "prj_8k2" })
← { auditId: "aud_4f01", status: "queued" }
→ get_audit({ auditId: "aud_4f01" })
← { status: "completed", findings: [ ...212 ] }
- Connect directly as a remote Streamable HTTP MCP server — no install step
- Stdio-only clients (e.g. Claude Desktop) bridge in via the og-lint CLI's mcp command
- Every tool call is workspace-scoped and metered exactly like a dashboard action
API & MCP, answered
Full CRUD over projects, templates (including publish), sharing links (including analytics), and audits (including triggering a crawl). It's the same operations the dashboard performs, scoped to whichever workspace the API key belongs to.
Generate an API key (og_live_...) in Settings > API keys — only a hash is ever stored, the plaintext is shown once. Send it as Authorization: Bearer og_live_... on every REST or MCP request.
Same operations, two transports. The REST API is a conventional JSON API for scripts, CI, and backends. The MCP server exposes the same operations as tools an AI agent can call directly — connect it as a remote MCP server, or bridge it to a stdio-only client with the og-lint CLI's mcp command.
No. Triggering an audit returns immediately with the audit in "queued" status — the crawl runs in the background. Poll GET /api/v1/audits/:id (or call the get_audit MCP tool) for status and findings as they fill in.
Yes, the same pool as the dashboard — generations, AI credits, audits, and active links are metered identically no matter which surface triggered them.
Stop paying per seat
for a usage-shaped problem.
Unlimited teammates, one usage pool. Start free with the scanner — no card required.