@browserid-ng/wallet

Give your AI agent its own identity. This is an MCP server that lets an agent provision a browserid-ng identity, get human-approved warrants, and present verifiable assertions — so it can sign in to services as itself, acting for you, within scopes you approve and can revoke.

No checkout, no build. Add one line to your MCP client and you're done.

Install

Add to your MCP client config (Claude Code / Cursor .mcp.json, or Claude Desktop's config):

{
  "mcpServers": {
    "browserid": { "command": "npx", "args": ["-y", "@browserid-ng/wallet"] }
  }
}

Then, in your agent, try the demo:

Provision a browserid-ng identity and sign the guestbook saying "hello from my agent".

The agent shows you an approval link (open it, confirm the fingerprint, approve), then signs the public guestbook at browserid.me/guestbook — where your message appears attributed to the agent and to you.

Tools

• provision(handles?, label?) — pair a new identity. Returns an approval URL for you; the agent picks up its identity automatically once you approve. No file to download — the private key is generated locally and never transmitted.
• identity — who the agent acts as.
• authorize(audience, scopes) — request a warrant for an audience.
• get_assertion(audience) — a backed assertion to present there.
• sign_guestbook(message) / read_guestbook() — the demo.

Where the identity lives

In ~/.browserid/ (the provisioning key, delegation, and cert). It's local to your machine — browserid.me never holds it.

Config

• BROWSERID_BROKER — default https://browserid.me.
• BROWSERID_HOME — default ~/.browserid.
• GUESTBOOK_URL — default <broker>/guestbook.
• AGENT_NAME — pick a reserved name for a multi-name credential.

License

MPL-2.0