MCP

Ask your own AI
about your own data.

Connect Clarvera to the AI assistant you already use. Ask about credentialing in plain language instead of clicking through screens. Your assistant does the thinking on your own subscription; Clarvera supplies the data, under exactly the permissions you already have.

Before you start

You need two things:

  1. A Clarvera login: the same details you use for the console.
  2. A plan that includes AI: Intelligence or above.

On a Core plan the connection still succeeds and you will see no tools. That is the plan limit working, not a fault.

Claude Code

Run this once. It applies to every project on your machine. 

claude mcp add --transport http --scope user \
  --client-id clarvera-mcp-client \
  --callback-port 3118 \
  clarvera-mcp https://val-mcp.clarvera.shinzox.com/mcp

Then start Claude Code, type /mcp, choose clarvera-mcp, and choose Authenticate. Your browser opens the Clarvera sign-in page; sign in and the connection completes.

All three options matter:

--scope user
Makes the connection available in every folder. Without it, it works only in the folder you ran the command in.
--client-id
Identifies Claude to Clarvera's sign-in service. Without it, sign-in is refused.
--callback-port
The address your browser returns to after sign-in. Without it a random port is used, and sign-in fails.

The client id is not a password. It is safe to share and to put in a shared setup script.

Claude in VS Code

Nothing extra to do. The VS Code extension reads the same configuration as the command line, so once you have run the command above, Clarvera is available there too. Reload the window if VS Code is already open.

Claude on the web

Not available yet. Claude's sign-in address has to be registered with Clarvera first. Until then, use Claude Code or VS Code above.

What you can ask

Start with “what tools do you have from Clarvera?”. Your assistant will list what your plan and permissions allow. Three areas are available today:

Providers

Credentialing status, licences, expiring credentials, incomplete records

Compliance

Exclusion checks and their results

Verification

Primary source verification requests and their status

Looking things up is switched on. Making changes is not. Clarvera can also let an assistant update records, and we have deliberately turned that off for launch so that every change still passes through a person. Until we switch it on, your assistant answers questions and updates stay in the console.

If it does not work

Connected, but no tools

Your session started before you connected. Close Claude Code and start it again.

Still no tools after restarting

Your plan does not include AI. Ask your Clarvera administrator about Intelligence.

The server is not listed at all

It was added to one folder only. Run the setup command again, including --scope user.

Sign-in fails

Not a Clarvera account, or the wrong one. Check the same details sign you in to the Clarvera console.

Dynamic Client Registration rejected

--client-id was missing. Remove the connection and add it again with all three options.

Still stuck? Send support the exact message you saw. The wording tells us which step failed.

Privacy and auditing

Every request is tied to your Clarvera account and limited to your organisation's data. The same row-level rules that apply in the console apply here, and every lookup is recorded in the audit trail against your account.

Your questions and your assistant's replies are handled by your own Claude subscription under Anthropic's terms. Clarvera sees the data requests that reach us, not your conversation.