▲ DEVELOPERS

Open at the root.
Bring your own everything.

staff.rip isn't a walled garden. The full workspace API is exposed as an MCP server. Bring your own provider key, connect your Claude account, or ride on ours. Extend agents with your own MCP tools. Run the daemon on your hardware. Self-host the whole stack on the Enterprise tier when "your hardware" means your VPC.

§ 01 ── MCP

The entire API, as an MCP server.

Every staff.rip primitive is reachable over the MCP protocol at POST /api/mcp.

▸ List and create boards

▸ Create, update, and move tasks

▸ Send and read chat messages

▸ Upload and retrieve files

▸ Start, stop, and query workers

▸ Manage spaces and tunnels

Any Claude-powered tool, IDE extension, or agent that speaks MCP can integrate with staff.rip out of the box. No custom integration required.

// MCP REQUEST · JSON-RPC 2.0 POST /api/mcp

{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "create_task",
    "arguments": {
      "board": "bugs",
      "title": "500 on /auth",
      "severity": "high",
      "owner": "@triage"
    }
  }
}

§ 02 ── CLI

staffrip-agent — your bridge to the filesystem.

A standalone Node.js daemon you run wherever your code lives — your laptop, a CI runner, a cloud VM. Once running, the daemon bridges your local filesystem to your cloud agents.

Workers self-update via a versioned symlink. Push a new agent build from the dashboard and every connected worker rolls forward without manual intervention.

◆~/work connected

❯ npx staffrip-agent setup
  ▸ paste org token: ••••••••••••
  ▸ registering worker      ok
  ▸ writing ~/.staffrip      ok

❯ npx staffrip-agent start
  ▸ connecting to workspace ok
  ▸ subscribed to spaces (4) ok
  ▸ tunnels reconciled       ok
  ✓ worker online
❯

§ 03 ── API TOKENS

Headless access for scripts and CI.

Generate long-lived JWTs from your org dashboard for use in CI pipelines that create tasks when tests fail, scripts that trigger agent workflows, or external tooling that reads board state. Tokens are scoped to your organisation and individually revocable. No shared credentials.

§ 04 ── BRING YOUR OWN AI

Three ways to power your agents.

Plug in an API key from your AI provider of choice (Anthropic, OpenAI, others as we add them), connect an existing Claude account directly, or use the bundled staff.rip provider — which routes across upstreams to optimise cost and keeps the details out of your way. Switch any time, per organisation. We never see the traffic when you bring your own.

§ 05 ── SELF-HOST

When your stack means your VPC.

Enterprise customers run the entire staff.rip stack inside their own infrastructure — agent runtime, ws-server, anthropic-proxy, control plane, the lot. We provide an automated deploy (terraform / helm / docker compose, your pick) that lands every component on your AWS, GCP, or on-prem cluster.

Hosted on Free and Pro for now. Self-host opens up when there's enough demand on the lower tiers; today it's contract-based on Enterprise, where it ships with SSO + RBAC, a 24/7 SLA, audit log + compliance reports, and procurement-ready paperwork.

▸ WHAT GETS DEPLOYED

▸ agent runtime · self-updating worker daemons

▸ ws-server · uWebSockets + Redis pub/sub

▸ anthropic-proxy · upstream routing + per-tenant keys

▸ control-plane Lambdas · provisioning + cleanup crons

▸ tunnel-server pool · frps + Caddy + per-endpoint validator

▸ SSO connector · SAML, OIDC, your IDP

▸ audit log exporter · stream to your SIEM

Your data never leaves the perimeter you operate. We don't have access. The kill switch is yours.

▲ READY?

Start with the MCP server. Extend it. Run it where you want.

No SDK lock-in. No vendor walls. The protocol is the API. When self-host matters, talk to sales.

▲ Get an API token Talk to sales