API & Integrations
Connect Claude and ChatGPT to your audits
Let an AI assistant read your audits directly and answer follow-up questions, instead of exporting a file each time.
Last updated 2026-08-25
In the app
Who this is for
Anyone who wants to ask questions about their audits in plain English: "which client site is worst on security?", "what should we fix first on brushbaby?", "has columbiagroup improved since June?".
You do not need to be technical. The browser route below is three clicks and involves no keys, no files and no code.
Needs a Pro or Enterprise plan.
Three ways in, and which to pick
| Best for | What it costs you | |
|---|---|---|
| Connect Claude (below) | Asking questions, following up, exploring | Three clicks, once |
| Build a ChatGPT action (below) | The same, if you use ChatGPT rather than Claude | About five minutes, once |
| Export a file | Handing a fixed snapshot to someone, or a spreadsheet | A download each time the data changes |
Either connection reads live data, so an answer is never out of date. An export is a photograph: fine for sharing, stale the moment a new audit finishes.
Connect from claude.ai
- In Claude, open Settings, then Connectors.
- Choose Add custom connector.
- Paste this address and confirm:
https://api.reviewmysite.app/mcp
- Claude sends you to ReviewMySite to sign in and approve. Press Allow.
That is all. Ask it something: "Which of my sites has the worst security score, and why?"
If your workspace uses your own branded address, use that instead, for example https://dashboard.youragency.com/mcp. Everything stays on your own domain.
Connect from Claude Code, Cursor, or your own tooling
Anything that can send an Authorization header can connect with an API key.
- Create a key under API Keys with the audits:read scope.
- Add it:
claude mcp add --transport http reviewmysite https://api.reviewmysite.app/mcp \
--header "Authorization: Bearer YOUR_API_KEY"
Connect from ChatGPT
ChatGPT does not offer the same one-click connector as Claude on every plan, so the route here is a custom GPT with an action. It reads the same live data and takes about five minutes.
You need a plan that can create a GPT (Plus, Pro, Team or Enterprise on the ChatGPT side) and a Pro or Enterprise ReviewMySite workspace.
- In ReviewMySite, open API Keys and create a key. Tick audits:read and sites:read. Copy it now: it is shown once.
- In ChatGPT, open My GPTs, then Create, then the Configure tab.
- Under Actions, choose Create new action, then Import from URL, and paste:
https://api.reviewmysite.app/api/openapi-gpt.json
- Set Authentication to API Key. Choose Custom for the header, set the header name to
X-API-Key, and paste your key as the value. - Save, then ask it something: "List my sites, then tell me which has the worst security score."
If your workspace uses your own branded address, use that host instead, for example https://dashboard.youragency.com/api/openapi-gpt.json.
The two things that catch people out
Use the X-API-Key header name. ChatGPT offers Bearer as the default auth type and it does work, but only if you leave the token exactly as issued: keys begin aud_. If you are unsure, choose Custom and X-API-Key, which is the documented header.
You do not need a workspace or tenant ID. The key already identifies your workspace, so there is nothing to look up and no x-tenant-id header to add. The Workspace ID shown under Settings > Organization is only for integrations that sign in as a user instead of using a key.
If a call comes back 401 or 403
| What you see | What it means |
|---|---|
401 Authentication required | The key did not reach us. Check the header name is X-API-Key, and that you pasted the whole key including the aud_ prefix |
401 Invalid API key | The key is wrong, revoked, or expired. Create a fresh one under API Keys |
403 with PLAN_FEATURE_UNAVAILABLE | The workspace is on Free. Programmatic access needs Pro or Enterprise |
403 Insufficient permissions | The key is missing the scope that endpoint needs. audits:read and sites:read cover the read actions |
The schema at the address above is deliberately read-only, so a GPT holding your key can look at audits, sites, reports and findings, and cannot start, change or delete anything.
What it can answer
| Ask about | It uses |
|---|---|
| What we track and how each site is doing | list_sites |
| Which site is worst, what ran this month | list_audits |
| Everything behind one audit | get_audit |
| What is actually wrong with a site | get_audit_report |
| Whether a site has improved | get_site_trends |
| What to fix first, and whether it is worth it | list_opportunities |
| Which page is slowest or most broken | list_pages, get_page |
| Brand, claims, AI citations, customer voice | get_intelligence |
Reading, and changing
When you approve a connection there is a box, Also let it make changes. It starts unticked.
Left unticked, the assistant reads and does nothing else. It cannot start an audit, add a site, or alter anything.
Ticked, it can also start audits and add sites. Two limits hold regardless:
- It uses your plan's monthly audit allowance. If that is used up, it refuses and tells you rather than spending your prepaid report credits. Buying credits stays a decision you make deliberately.
- Nothing can delete anything, at any setting.
An audit started this way runs in the background and takes around eight minutes, sometimes longer on a large site. The assistant gets an id back and can check on it.
What it can and cannot see
A connection is bound to one workspace, the one you chose when approving it. If you belong to several, connect each one separately.
It sees exactly what you see. A team member limited to one client group gets a connection limited to that same client group.
Every tool call is written to your Audit log, with which tool ran and the filters it used, so you can always see what was read.
Exporting a file instead
On Reports and on Audits, Export audits downloads a batch as one file. Markdown is the one to hand to an assistant that is not connected; CSV suits a spreadsheet. This button works on every plan, including Free. If you want the assistant to read live data instead of a snapshot, connect it using one of the routes above.
Full detail is in Sites and reports.
If something does not work
- "Needs a Pro or Enterprise plan": programmatic access, which includes assistant connections, is a paid feature. Upgrade under Billing.
- The assistant says it cannot start audits: the write box was left unticked. Reconnect and tick it.
- It says the allowance is used up: that is the credit guard working. Start the audit yourself from the dashboard if you want to spend a credit.
- Answers look thin on brand or citations: the AI intelligence pipeline is off by default and produces that data. The reply says so rather than implying the site is clean.
- A long report comes back cut short: reports can be very large and have to fit in the assistant's context. The reply says it was truncated and links the full version.