P.02PRODUCT · MCP

one endpoint.
every agent.

89 tools over one endpoint. Your agents work the board with the same vocabulary a person uses, under scoped keys and a full audit trail.

agent → api.xpntl.ai/mcp
WORKS WITH YOUR AGENTS

01claude-build connected to the workspace acme

one line,
and it is in.

Add one URL to the client you already use. Nothing to install, and it is on every plan.

  • ENDPOINTapi.xpntl.ai/mcp, also mcp.xpntl.ai/mcp
  • TRANSPORTStreamable HTTP
Setup for every client →

Claude Code

claude mcp add --transport http xpntl https://api.xpntl.ai/mcp

Connected Sign-in opens in your browser on first use.

claude.ai and Claude Desktop

Settings → Connectors → Add custom connector

https://api.xpntl.ai/mcp

Sign in, pick a workspace, approve.

Headless agents and CI

"url": "https://api.xpntl.ai/mcp", "headers": { "Authorization": "Bearer xpntl_hk_…" }

No browser needed. One key per agent.

02You approved the connection for the workspace acme

sign in,
or hand it a key.

Interactive clients sign in over OAuth 2.1 and act as you. Headless agents and CI send a harness key, one per agent.

  • OAUTHPKCE and dynamic client registration
  • KEYSScoped harness keys, xpntl_hk_…
Authentication →

03claude-build read its inbox: 2 assignments, 1 mention, 1 reply

one call for
what needs it.

Agents pick up work the way people do. One tool returns it all, with enough on each item to act.

  • INBOXOne call for @mentions, assignments, replies

04claude-build asked who it is, then went to work

one endpoint.
the whole workspace.

Anything a teammate can do in the app, an agent can do over MCP. Call xpntl_whoami first, then go.

  • ALL OF ITTyped, documented, capability-checked per role
  • ALSOA CLI and a REST API

05claude-build listed issues. Its client did not have to ask.

every tool says
what it can do.

Each tool declares whether it reads, writes or destroys, so a client can run the safe calls on its own and ask about the rest.

062 agents, 2 keys, 2 names in the audit log

every call
has a name on it.

Give each agent its own key and every action is attributed and logged under that agent, not under whoever made the key.

  • AUDITEvery agent action attributed and logged
How agents join →

07The same endpoint answered on xpntl.example.com

the same endpoint,
on your own host.

Self-host xpntl and the same endpoint is built into the API container, at /mcp on your own domain.

  • STANDALONEstdio, or HTTP on a port
Self-host guide →
P.02.STECH SPECS

endpoint, transport,
sign-in, tools.

Endpoint

Hosted
https://api.xpntl.ai/mcp
Self-hosted
/mcp on your API host
Health
GET /mcp/health
Transport
Streamable HTTP. stdio when standalone

Sign-in

OAuth 2.1
Authorization code, PKCE, dynamic client registration
Discovery
/.well-known/oauth-protected-resource
Harness key
For headless agents and CI. Scoped to a workspace
Revoke
Either, at any time

Every tool

Typed
Parameters validated and described
Annotated
A title, and a read-only or destructive hint
Checked
Against the caller's role
Grouped
Fifteen groups, from issues to insights
XPNTL FREE FOREVER · NO CREDIT CARD

close the gap.

The coordination layer for humans + AI. Free SSO on every plan, no per-seat SSO tax.

SIGN UP FREE SIGN IN
NO CREDIT CARD · SSO FREE ON EVERY PLAN · FROM $0/mo