MCP for Claude Module Included
The simplest of the three. Claude registers itself — there is no client to create, no secret to paste. A user adds the connector, signs in to Vaks in the browser, and is working. This is the shortest guide for that reason.
Before you start
Everything Claude-specific is on this page. The shared parts live in the hub:
- Enable the connector for your organization — Enable and scope it. Off by default; consent is refused until it is on.
- Permissions model — an assistant's rights are the user's rights. Read-only until you turn writes on.
- The URL to hand out — the tenant's MCP URL from the admin panel, or the public MCP URL if a gateway was deployed (reaching it from outside).
Connect Claude
Claude self-registers via Dynamic Client Registration — no pre-registered client, no secret. Each user does this once:
- In Claude (claude.ai or Claude Desktop), add Vaks PM as a custom connector.
- Paste the MCP server URL (
https://<tenant>/mcp, or your public MCP URL). - Claude discovers how to authenticate and opens a browser sign-in to Vaks (local or SSO).
- The user approves the consent screen and Claude starts working — standard OAuth with PKCE, one-hour tokens that refresh automatically.
Per-user by nature. Each Claude user signs in as themselves, so the assistant acts as that person, bounded by their own permissions. In Vaks, Admin → Integrations → MCP → Active AI connections shows one connection per user.
Claude Desktop and IDE clients too. Desktop and CLI clients that redirect to a local loopback port (the varying port they pick each launch) are recognised as such and accepted automatically — the same self-registration path, no extra configuration.
Pre-built workflows
Good news: Claude surfaces the Vaks pre-built workflows automatically. The MCP prompts appear as a menu of one-click slash-commands (Portfolio · Risk radar, Project · Status report, Team · Staff a project, and so on) — nothing to recreate, unlike Copilot Studio and ChatGPT. See the full library in the hub.
Write workflows appear only when writes are enabled. Read workflows (risk radar, status report, find work, budget audit) are always listed. The ones that end in a write — staff a project, close out, announce a change, onboard someone, timesheet catch-up, start from a brief — appear only once Allow AI write actions is on. A user who was connected before you turned writes on may need to reconnect to see them.
Troubleshooting
| Symptom | Cause & fix |
|---|---|
| Claude cannot reach the server | Almost always the URL: on a gateway deployment the admin-panel URL is not the public one. Confirm the public hostname — see reaching it from outside. Claude is a cloud service, so the endpoint must be reachable over the internet. |
| Consent screen refuses, no approve button | The connector is disabled. Enable it in the admin section. |
| Only read tools / no workflow menu writes | Writes are off, or the connection predates turning them on. Turn on Allow AI write actions and reconnect. |
| New tools or prompts don't show up | An already-connected client picks up new tools on its next refresh — reconnect the connector to see them immediately. |
Related: MCP overview (hub) — permissions, admin toggles, deployment · Copilot Studio · ChatGPT · all integrations.