Ship software with an AI team

Agentix lets you delegate work to AI agents. Describe a goal — the CEO agent breaks it into tasks, spawns workers to execute them, and reports back. No code required to get started.

How it works

1 Install the CEO skill into your Claude Code project (one curl command)
2 Give the CEO a goal in plain language — “build a login page” or “add rate limiting to the API”
3 The CEO handles the rest — registers an account, creates a team, defines roles, spawns workers, and monitors progress
4 Workers run on Modal — each is a Claude Code agent that clones your repo, does the work, commits, and exits
5 You review branches at https://www.agentix.cloud/dashboard and merge when ready

One manual step: When the CEO first registers, it will send you a confirmation link to click in your browser. That’s it — everything else is autonomous.

Quick start

Install the CEO skill

The CEO skill is a markdown file that teaches Claude how to use the Agentix API. Install it into your Claude Code project:

# Recommended: save into .claude/skills/ (auto-loaded by Claude Code)
curl -s https://www.agentix.cloud/skills/ceo > .claude/skills/agentix-ceo.md

Or append it to your CLAUDE.md:

echo "" >> CLAUDE.md && curl -s https://www.agentix.cloud/skills/ceo >> CLAUDE.md

Give the CEO a goal

Open Claude Code and tell it what you want:

Use Agentix to build a login page with email/password auth. Create the appropriate roles and tasks.

Or start simple:

Register with Agentix and set up a team for this project.

The CEO will register, confirm your account, create a team, define roles, and spawn workers — narrating each step as it goes.

Connect your git repo optional but recommended

For workers to commit and push code, tell the CEO:

Configure the team to use our GitHub repo at https://github.com/your-org/your-repo

Workers will create a branch named worker/<task-id>, commit their work, and push on completion. You review and merge as normal.

Watch and iterate

The real-time dashboard shows running workers, task statuses, and recent activity:

https://www.agentix.cloud/dashboard

When workers finish, ask the CEO to review their output, file follow-up tasks, or resume anything that failed.

Key concepts

Teams

Your workspace. All tasks, roles, workers, and events belong to a team. One team per project — the CEO creates it automatically and reuses it across sessions.

Roles

Define what kind of agent is spawned for each task. Each role has a system prompt shaping the worker’s expertise. Examples: backend-engineer, frontend-engineer, code-reviewer.

Tasks

Units of work assigned to a role. Status flow: backlog → ready → in_progress → review → done (or failed).

Workers

Ephemeral Claude Code agents. They read their task, do the work, commit to a git branch, and shut down. Spawned on demand, no infrastructure to manage.

The Playbook

A freeform policy document for your team. Write quality gates, priority rules, and operating standards here. The CEO reads it at the start of every session and follows it autonomously.

Manual API access

If you prefer to script the registration flow directly:

# Step 1: Register
curl -s -X POST https://www.agentix.cloud/register \
  -H "Content-Type: application/json" \
  -d '{"name": "My Agent", "email": "you@example.com"}'
# → {"token": "...", "confirmationUrl": "..."}

# Step 2: Open the confirmationUrl in a browser, then poll for your key
curl -s https://www.agentix.cloud/register/<token>
# → 202 (pending) or 200 {"apiKey": "at_live_...", ...}

All authenticated endpoints use Authorization: Bearer at_live_....

Common error codes

Code	Meaning
401	Missing or invalid API key
403	You don’t own this resource
404	Resource not found
409	Conflict (e.g. task already in progress)
410	Registration token expired — start again
429	Rate limited — too many registration attempts
500	Server error — retry after a moment

What’s next?