Every agent connects the same way: point it at your dhvati endpoint. Your browser opens once to sign you in — after that, every tool call is verified, decided by policy, and audited.
https://mcp.dhvati.ai/mcp
One URL for everyone — who you are is decided at sign-in, not by the link.
1
Open connector settings
Settings → Connectors → Add custom connector
2
Name it and paste the endpoint
Name: dhvati · URL:
https://mcp.dhvati.ai/mcp
3
Sign in when the browser opens
Claude asks you to connect — your browser opens dhvati's sign-in. New here? Register creates your free workspace on the spot. Company account? Use your company's button on the same screen.
4
Ask for something
Your governed tools appear under Search & tools. Reads flow instantly; writes show an approval card. Every call lands in your console's audit trail.
1
Open connector settings
On claude.ai: Settings → Connectors → Add custom connector. On Team/Enterprise plans an admin can add it once for the whole workspace.
2
Paste the endpoint
https://mcp.dhvati.ai/mcp
3
Connect & sign in
Click Connect next to dhvati — sign in (or Register free) in the window that opens. Done: your tools are available in every chat where you enable them.
1
Add the server
In any terminal:
claude mcp add --transport http dhvati https://mcp.dhvati.ai/mcp
2
Authenticate
Inside Claude Code, run /mcp and pick dhvati — your browser opens to sign in (or register free).
3
Use it
dhvati's governed tools join Claude Code's toolbox. Denied calls come back with the precise policy reason — and a request your admin (or you, on the free tier) can approve from the console.
1
Add dhvati to your MCP config
Merge this into ~/.cursor/mcp.json (or the project's .cursor/mcp.json):
In Cursor: Settings → MCP — dhvati appears in the list. The first use opens your browser to sign in (or register free).
1
Point it at the endpoint
dhvati speaks MCP over Streamable HTTP:
https://mcp.dhvati.ai/mcp
2
Auth is standard OAuth — no keys to paste
Spec-compliant clients discover everything themselves: protected-resource metadata (RFC 9728), authorization-server metadata (RFC 8414), and dynamic client registration (RFC 7591) with PKCE. If your client supports MCP authorization, it will Just Work.
First use
Browser opens once — sign in or register. That's your identity being verified; no API keys, nothing pasted into the agent.
Every call after
Verified → checked against policy → allowed with a vault credential, or blocked with the reason. Reads flow; writes ask first.
In the console
Approve requests with one tap, connect more tools, invite teammates, and read the audit trail at my.dhvati.ai.
Setup details, permissions, and troubleshooting live in the docs. The dhvati connector is also headed to the Claude connectors directory — manual setup works today and won't change.
Want dhvati for your team?
Book a 30-minute walkthrough and we'll set it up on your use case. Building agents with the dhvati SDK? Join the developer waitlist.