mycel
Tutorial: Create Your First Agent
DocsGetting StartedTutorial: Create Your First Agent

Tutorial: Create Your First Agent

This tutorial walks you through creating, running, and communicating with your first AI agent in bc.

Prerequisites

  • A bc workspace initialized (bc init completed)
  • The root agent running (bc up)
  • An AI provider configured (e.g., Claude Code or Gemini)

Step 1: Create an agent

Create an engineer agent named eng-01:

bc agent create eng-01 --role engineer

This creates:

  • A git worktree at .bc/agents/eng-01/worktree/
  • A tmux session (or Docker container) for the agent
  • Role-specific configuration files (CLAUDE.md, .mcp.json)

Step 2: Verify the agent is running

bc status

You should see eng-01 listed with state idle:

AGENT    ROLE      STATE   UPTIME   TASK
root     root      idle    5m
eng-01   engineer  idle    10s

Step 3: Send work to the agent

bc agent send eng-01 "Add a health check endpoint that returns JSON with status and uptime"

The agent receives the message in its tmux session and begins working.

Step 4: Monitor progress

Watch the agent's output in real time:

bc agent peek eng-01

Or attach directly to the agent's tmux session:

bc agent attach eng-01

Press Ctrl+B then D to detach from the session without stopping the agent.

Step 5: Subscribe to notifications

Subscribe the agent to a notification source so it receives platform events:

# List available notification sources
bc notify list

# Subscribe eng-01 to a Slack channel
bc notify subscribe slack:engineering eng-01

# Check recent notifications
bc notify history slack:engineering --last 10

Step 6: Check costs

After the agent has been working for a while, check spending:

bc cost show
bc cost agent

Step 7: Stop the agent

When the work is complete:

bc agent stop eng-01

To clean up entirely (removes worktree and state):

bc agent delete eng-01

Next steps