Connect guide

Connect your agent in five steps.

The CLI path. Uses @teamx-sh/agent to watch your local git activity and report metadata back to the team dashboard. Takes about five minutes. Metadata only — your code never leaves your machine.

What the agent does

A small watcher. A big view.

Agent-aware

Detects Claude Code, Cursor, Codex, OpenClaw sessions and attributes activity to the right agent.

Auto-visible

Commits, task transitions, and session summaries stream into the team dashboard automatically.

Real-time

Your teammates see what every agent is building as it happens. No standups. No stale tickets.

Setup

Five steps

1

Sign up and create a team

Head to the dashboard, sign up with email or Google, create a team (or join one via invite link). 30 seconds tops.

2

Install the CLI globally

Install the TeamX agent CLI via npm. Requires Node.js 18+. Add sudo on Linux/macOS if you get a permissions error.

npm install -g @teamx-sh/agent
3

Connect to your team

Open Dashboard → Connect Agent and click the Agent CLI tab. TeamX mints a secure, member-scoped token — copy the connect command shown there and run it once in any terminal.

teamx-agent connect --token=<your-token>
4

Register as a startup service

One command registers the agent as a startup entry. No admin rights needed on any platform. It starts immediately and auto-restarts at every login — you never have to think about it again.

# Install (starts now + auto-starts at every login)
teamx-agent install-service

# To remove:
teamx-agent uninstall-service
5

Verify it’s live

Open Dashboard → Agents. Your machine appears as a live relay agent within about 30 seconds. Activity, commits, and session summaries start flowing to your team automatically.

What a connected agent unlocks

Connecting is just 5 minutes. What lands automatically after that is the interesting part.

Auto Docs on every release

Your laptop generates a tech doc + user doc on every epic close and feature-version ship — for free, no BYOK key required. See the Auto Docs guide.

Smart briefings on begin_task

Every task pickup fires a context briefing the agent reads first. Docs, decisions, recent commits, files likely to be touched. Context Engine.

Blast Radius on every PR

Your PRs get an auto-posted Affects: block and owners auto-requested as reviewers, scoped to your team's switch. Blast Radius.

Remote DM control

DM your local Claude Code from TeamX Messages and replies stream back inline. Remote control.

Privacy & security

TeamX is a metadata layer. The agent reads git history and session summaries — never your source code, never your environment variables, never your secrets. Tokens are scoped to a single team member and revokable instantly from the dashboard. All transport is HTTPS. Your data stays yours.

FAQ

What data does the agent send?
Metadata only. Commit messages, task status transitions, session durations, file-change summaries. Never your source code, never env vars, never secrets.
Is my code safe?
Yes. The agent reads git metadata and session info. No file contents or diffs are transmitted. All traffic is HTTPS.
Can I disconnect?
Anytime. Revoke the token from Dashboard → Connect Agent (instant invalidation), or kill the CLI process in your terminal. Revoke from the dashboard to be safe.
Can I track multiple projects?
Yes. Run the connect command in each project directory. Each gets its own agent process; all activity flows to the same team dashboard.
Which agents are supported?
Claude Code, Cursor, Codex, GitHub Copilot, OpenClaw — all via the MCP integration. Any agent that produces git activity can be tracked through the GitHub webhook path instead.
Does it work with CI/CD?
The CLI agent is for local development. For CI/CD visibility, TeamX connects directly to your GitHub webhooks to track deployments and pipeline runs.

Five steps. Let’s go.