deep-agent-cli

Deep Agents CLI

Build a terminal-based coding agent similar to Claude Code using the Deep Agents CLI or by creating your own CLI interface.

When to Use

• When you want a Claude Code-like experience in the terminal
• When building a CLI tool for AI-assisted coding
• When running Deep Agents interactively

Available Operations

1. Install and use the official Deep Agents CLI
2. Build a custom CLI with rich terminal UI
3. Configure CLI with local file access + shell
4. Add skills and AGENTS.md to CLI agent

Multi-Step Workflow

Step 1: Install Deep Agents CLI

pip install deepagents
# or
uv tool install deepagents

Step 2: Run the CLI

# Start interactive session
deepagents

# Start with a specific model
deepagents --model anthropic:claude-sonnet-4-6

# Start in a specific directory
deepagents --dir ./my-project

# Start with a prompt
deepagents "Fix the failing tests in src/auth/"

Step 3: Build Custom CLI

# cli.py
import asyncio
import sys
from deepagents import create_deep_agent
from deepagents.backends import LocalShellBackend
from langgraph.checkpoint.memory import MemorySaver

async def main():
    # Full coding agent with file + shell access
    agent = create_deep_agent(
        model="anthropic:claude-sonnet-4-6",
        backend=LocalShellBackend(root_dir="."),
        skills=["./skills/"],
        memory=["/AGENTS.md"],
        checkpointer=MemorySaver(),
        system_prompt="You are a coding assistant. Follow the project conventions."
    )

    config = {"configurable": {"thread_id": "cli-session"}}

    # One-shot mode
    if len(sys.argv) > 1:
        prompt = " ".join(sys.argv[1:])
        async for event in agent.astream_events(
            {"messages": [{"role": "user", "content": prompt}]},
            config=config, version="v2"
        ):
            if event["event"] == "on_chat_model_stream":
                print(event["data"]["chunk"].content, end="", flush=True)
        print()
        return

    # Interactive mode
    print("Coding Agent (type 'exit' to quit)\n")
    while True:
        try:
            user_input = input("> ")
        except (EOFError, KeyboardInterrupt):
            break

        if user_input.lower() in ("exit", "quit"):
            break

        async for event in agent.astream_events(
            {"messages": [{"role": "user", "content": user_input}]},
            config=config, version="v2"
        ):
            if event["event"] == "on_chat_model_stream":
                print(event["data"]["chunk"].content, end="", flush=True)
        print("\n")

asyncio.run(main())