MCP integration — connect AI agents to your data
Supasaito exposes a single endpoint that any MCP-compatible AI client or agent can use to securely read — and, with permission, act on — your AI-visibility data. It inherits the exact same permissions and tenant isolation as the dashboard.
Set permissions for each connection
- PATs and OAuth connections belong to a user, whose full name is shown. Choose the companies and own brands each connection may access.
- Writing, deletion, creating companies and creating campaigns have separate permissions. Set brand creation per company. Including future companies or brands requires an explicit choice.
- Edit an existing connection in Settings → MCP & API tokens. Older connections keep their previous scope until you review and save their permissions.
- In History & Access, company Owners and Admins can block one PAT, one OAuth app, or all current and future MCP connections of a member. Other companies and normal console access are unaffected.
What it can do
- Read the 5-axis analytics (visibility, position, sentiment, competitors, sources) in every cut.
- Read recommendations (content, context, technical), change their status, and generate content briefs and rewrites.
- Run campaigns and tasks: an agent takes a task from the queue, writes brand and campaign documents with revision history, publishes a placement and submits the work for review — under the same review rules you set.
- Create and update brands / topics / prompts / competitors / sources (with a read+write token).
- Browse past-scan history and run manual scans.
- Read how the numbers move scan over scan — you against competitors, by topic, by assistant — with the scans whose measurement changed flagged.
Run the audit and fix the site
- Trigger a site audit and read each page’s readiness scores (SSPS, SSPS+, PERF) and the exact problems found.
- Pull every technical recommendation with its ready-to-apply brief — the precise robots, JSON-LD or meta change, with acceptance criteria.
- With write access (any paid plan), an agent can apply the fix, re-run the audit, and keep going until the issue resolves.
Safe by design
- Same permissions as the dashboard — row-level security (RLS) and your role (Owner / Admin / Member / Viewer) are inherited per call.
- Tokens run under a per-company rate limit, so one company’s usage never drains another’s.
- An owner/admin can revoke any token’s or app’s access to their company at any time, from Settings → History & Access.
How to connect
- Personal Access Token (PAT) — create one under Settings → MCP & API tokens and pass it as a Bearer token.
- OAuth 2.1 — hosted clients like claude.ai and ChatGPT connect automatically.
Plan your MCP integration
Book a free consultation to discuss your AI tools, workflow and setup. MCP is available on Starter and higher paid plans.
Book a free consultation