Integrations

Pick your runtime.

How it works

One endpoint. Every runtime.

AgentValet exposes a single MCP endpoint. Register once and any MCP-capable agent gets the same governance layer. The per-runtime guides differ only in which config file the CLI writes for you.

01
Register your agent
Run npx @agentvalet/register in your project. The CLI opens your browser, signs you in, generates an RS256 keypair, and stores it on this machine. No IDs to copy.
02
The CLI wires your runtime for you
It writes the right MCP config for whatever you're running: .mcp.json, .cursor/mcp.json, .factory/mcp.json, ~/.codex/config.toml, so you don't paste anything.
03
Approve platforms in the dashboard
Connect GitHub, Slack, HubSpot, or any of 20 live platforms in the AgentValet dashboard. Each platform gets a scoped permission matrix per agent.
04
Every call is logged and revocable
Credentials are decrypted in-memory at call time, never stored in responses or logs. Revoke any agent or platform instantly from the dashboard.
OAuth Connect

Or connect any MCP client over OAuth.

AgentValet is a native OAuth 2.1 Authorization Server with Dynamic Client Registration (RFC 7591) and PKCE. MCP clients that support browser-based OAuth consent connect directly, with no local key to manage. Clients that require Dynamic Client Registration, like Codex CLI, work too, because AgentValet implements it out of the box. The consent flow provisions the agent a signing identity, not a raw credential. Nothing long-lived sits on disk.

Browser consent, not a token file
Your MCP client opens a browser tab. You approve the connection. AgentValet provisions the agent its own signing identity and records the agent type in the audit log. No bearer token ends up in a config file.
DCR + PKCE, built in
AgentValet implements RFC 7591 Dynamic Client Registration and S256 PKCE. Clients that require DCR (like Codex) work without any pre-registration step on your part.
Same governance, different path
Whether you connect via the CLI or OAuth, the result is the same: the agent runs through your permission matrix, every call is logged, and you can revoke access from the dashboard at any time.

Quick reference: connect your client

Claude Code  Full guide →

Add as a remote MCP server, then run /mcp to trigger the OAuth browser flow. The token is stored in the OS keychain and auto-refreshed.

terminal
claude mcp add --transport http agentvalet https://api.agentvalet.ai/mcp/v1/<owner-id>
Claude Desktop  Full guide →

Add AgentValet as a remote MCP server in Settings → Connectors using the URL below. OAuth consent runs in your browser. No plan gating for OAuth connections.

remote MCP server URL
https://api.agentvalet.ai/mcp/v1/<owner-id>

Add the URL in Settings → MCP. Cursor shows a "Needs login" button in the MCP panel. Click it and OAuth consent opens in a browser tab.

remote MCP server URL
https://api.agentvalet.ai/mcp/v1/<owner-id>
Codex CLI  Full guide →

Codex's OAuth requires the server to implement Dynamic Client Registration. AgentValet does. Two commands: add the server, then login to open the browser flow.

terminal
codex mcp add agentvalet --url https://api.agentvalet.ai/mcp/v1/<owner-id>
codex mcp login agentvalet
Identity, not credentials. Whichever path you take, the agent is issued its own signing identity. No long-lived bearer token sits in a config file.

Ready to govern your agents?

Register in under 5 minutes. No credit card required.