Your consent platform, one prompt away.
Connect an assistant to the Katla account you already have. Ask what your sites are setting, run a scan, fix a classification — in the same window you were already working in.
Which of our sites dropped below 70% consent this week?
Rescan the shop and tell me what is new.
One URL. No install, no keys.
It is a remote server, so there is nothing to run locally and no API key to create. You sign in with the Katla account you already have.
- Add function(){throw Error("Attempted to call MCP_URL() from the server but MCP_URL is on the client. It's not possible to invoke a client function from the server, it can only be rendered as a Component or passed to props of a Client Component.")} to your client.
- Sign in with your Katla account when the browser opens.
- Ask. Write tools declare themselves, so your client can confirm first.
1. Open Settings → Connectors 2. Add a custom connector 3. Paste the server URL: https://api.katla.app/mcp 4. Click Add, then sign in when the browser opens.
Eighteen tools, and nothing hidden behind them.
The same operations the dashboard and the CLI use. Ten read, eight write.
The questions that used to need a ticket.
“Did anything new start setting cookies on staging this week?”
Runs a scan and diffs the inventory against what was there before.
“What is our accept rate, and where is it worst?”
Consent stats per site, so the banner copy conversation starts with numbers.
“Give me every cookie we declare as marketing, with its purpose.”
The classified inventory with the descriptions your policy publishes.
“Add these four domains and verify them.”
Adds each site and checks the DNS record or meta tag for it.
“Reclassify _hjSession as functional and say why.”
Writes the category and your own description, in the locales you serve.
“Regenerate the policy now that the scan is done.”
Publishes the policy from the inventory the scan actually found.
It can only do what you can.
Connecting an assistant should not widen anyone's access. It authenticates as you, over OAuth 2.1, and carries the scopes you granted it.
- Scoped
- Tokens carry katla:read or katla:write. A read-only connection cannot start a scan or change a classification.
- Acts as you
- Tools resolve through your own account, so an assistant reaches exactly the teams and sites you do — no more.
- Writes are declared
- Each tool declares whether it mutates anything, so your client can ask before it runs one.
- Revocable
- Authorisation is OAuth 2.1 and can be revoked at any time, which cuts the connection off immediately.
Paste one URL and ask it something.
https://api.katla.app/mcp