01
Connect in a minute
Add https://mcp.secani.com/mcp to Claude, ChatGPT, Cursor or Codex and sign in with your Secani account. No API key to manage.
OAuth sign-in through WorkOS; every token belongs to one user and one client.
Your ISMS in every AI tool. Every write confirmed. Every call checked.
Add one URL to your AI client and sign in. Your assistant can search controls, risks and evidence and propose changes, while Secani checks your access on every call.
One URL, every client.
Context flows to your assistant, and changes flow back after you confirm.
02
Ask with real context
Your assistant searches requirements, risks, measures, evidence and tasks directly, instead of working from what you pasted into the chat.
Reads cover search, entity details, risks, findings, measures, tasks, requirements, evidence, obligations and ISMS status.
03
Write after you confirm
When your assistant saves a measure, risk or evidence link, your client asks you first. Nothing is written without your yes.
Writes reuse Secani's own commands, are idempotent per request and fail instead of overwriting a newer change.
What stops being manual.
The copy and paste between your assistant and your ISMS.
Without SecaniWith the MCP server
Pasting controls and risks into a chat for context
The assistant reads them directly, with your access
Retyping the assistant's answer into Secani
Proposed through MCP, saved after your confirmation
Linking evidence to requirements by hand
Linked from your AI client in one confirmed step
Managing API keys for every tool
OAuth sign-in per user and client, revocable anytime
Count the pasting.
How often do you copy context into an AI chat and the answer back?
Hours per year spent copying
25
Your numbers, multiplied: rounds × minutes. Nothing of ours goes into the result.
Powerful, and on a short leash.
The assistant works with your access, never more.
Your access, every call
Each call is checked against your access to the organization, workspace and governance scope. The assistant cannot see more than you can.
Confirmation for every write
Writes ask you in the client first. If the client cannot ask, the call fails and nothing is written.
Revocable per client
See your connected clients, revoke one or switch it to read-only. Writes are recorded with the AI as actor and you as the person behind it.
Questions about MCP
Any client that supports remote MCP servers with OAuth. We document Claude Code, Codex CLI, Cursor and ChatGPT custom connectors.
No. Every write asks for your confirmation in the client. If the client cannot show the prompt, the write fails.
Exactly what your account may see, checked on every call against organization, workspace and governance scope.
Your connected clients are listed in Secani. Revoke one, restore it later, or set it to read-only so it can no longer write.
Stop pasting your ISMS into chats.
Connect your assistant and keep every write in your hands.