DocsCLIco ai

co ai

AI coding agent that works in your project. Start a web chat session or run one-shot prompts — with full access to your files, shell, and tools.

Quick Start: Run co ai in your project — it opens a web chat connected to a coding agent that can read and edit your files.

Two Modes

Web Chat (default)

co ai

Starts a local agent server, opens chat.openonion.ai in your browser. Chat conversationally — best for extended sessions.

One-Shot

co ai "your prompt"

Runs the prompt, prints the result, exits. Best for quick tasks and scripting.

Web Chat Mode

Run co ai in your project directory:

code

co ai

Python REPL

Interactive

Starting AI coding agent...

Server running at http://localhost:8000

Opening chat.openonion.ai/0x7a9f3b2c...

What happens

1.Agent server starts on localhost:8000

2.Your browser opens chat.openonion.ai/{your-address}

3.The agent runs in your project directory with full tool access

4.Your identity comes from ~/.co/ — same across all sessions

One-Shot Mode

Pass a prompt directly — the agent runs, prints the result, and exits:

code

co ai "Create a FastAPI hello world app"

Python REPL

Interactive

[agent] Creating FastAPI hello world app...

[agent] Writing main.py...

from fastapi import FastAPI

app = FastAPI()

@app.get("/")

def hello():

return {"message": "Hello World"}

Done! Run with: uvicorn main:app --reload

code

# Fix a bug
co ai "fix the failing test in tests/unit/test_agent.py"

# Add a feature
co ai "add rate limiting to the /api endpoint"

# Refactor
co ai "refactor agent.py to use the new event system"

# Write tests
co ai "write pytest tests for models/user.py"

Project Context

When started, the agent automatically loads context from your project — no setup required:

Source	Purpose
.co/OO.md	Project instructions (primary)
CLAUDE.md	Claude Code compatibility
README.md	Project overview
~/.claude/skills/	User-defined skills
git status	Current branch, changes, recent commits

Give the agent persistent project rules with .co/OO.md:

code

mkdir -p .co
cat > .co/OO.md << 'EOF'
Always run tests before committing.
Use snake_case for function names.
The main entry point is src/main.py.
EOF

Loaded every session — the agent always follows your rules without being told.

Available Tools

File Operations

• Read, write, edit files
• Search with glob and grep

Shell

• Run bash commands
• Approval flow for destructive ops

Planning

• Enter plan mode, write plans
• Exit plan and implement

Tasks & Skills

• Background tasks, todo lists
• Load and run user-defined skills

Command Reference

Option	Short	Default	Description
--port	-p	8000	Web server port
--model	-m	co/claude-opus-4-5	LLM model to use
--max-iterations	-i	100	Max tool iterations per turn

code

# Different port
co ai --port 9000

# Faster model
co ai --model co/gemini-2.5-pro

# One-shot with options
co ai "build a microservice" --model co/gpt-4o --max-iterations 50

CLI Reference

co browser

co ai

Two Modes

Web Chat (default)

One-Shot

Web Chat Mode

What happens

One-Shot Mode

Project Context

Available Tools

File Operations

Shell

Planning

Tasks & Skills

Command Reference

Enjoying ConnectOnion?