Deep Agents CLI: A Terminal-Based AI Coding Assistant

Mar 20, 2026

Purpose

I wanted an AI coding assistant that works in my terminal. I use Claude Code sometimes, but I wanted something open-source that works with different LLM providers. Deep Agents CLI seemed like a good option - it’s a pre-built terminal agent that supports multiple models.

What is Deep Agents CLI?

Deep Agents CLI is a terminal-based AI coding assistant. It’s similar to Claude Code or Cursor, but:

Works with any LLM that supports tool calling (not just Claude)
Open-source and extensible
Supports custom skills via slash commands
Can run in remote sandboxes

Environment

macOS or Linux
Terminal
An LLM API key (Anthropic, OpenAI, etc.)

How to Install

One-Line Install

The fastest way is using the install script:

curl -LsSf https://raw.githubusercontent.com/langchain-ai/deepagents/main/libs/cli/scripts/install.sh | bash

Or with uv

If you already have uv installed:

uv tool install 'deepagents-cli[nvidia,ollama]'

The extras in brackets add support for specific providers. Use nvidia for NVIDIA NIM, ollama for Ollama models.

Features I Found Useful

Interactive TUI

After installing, I run:

deepagents

This opens an interactive terminal interface with streaming responses. I can type prompts, see the agent think and act, and continue the conversation.

Conversation Resume

The CLI remembers conversations. When I start it again, I can pick up where I left off. This is useful for long coding sessions where I need to step away.

Web Search

The agent can search the web to ground its responses. This helps when I ask about recent library versions or current documentation.

Persistent Memory

The agent remembers context across conversations. If I tell it about my project structure, it recalls that information later.

Headless Mode

For CI/CD pipelines or scripts, I can run non-interactively:

deepagents --headless "Fix the linting errors in src/"

Human-in-the-Loop

I can configure the CLI to ask for approval before executing tool calls. This gives me control over what the agent does.

Custom Skills

I can extend the agent with custom slash commands. Skills are stored in a skills repository and can be loaded at runtime.

To use a skills repo:

deepagents --skills-repo myorg/skills-repo

GitHub Actions Integration

Deep Agents CLI includes a GitHub Action for automated workflows:

jobs:
  agent:
    runs-on: ubuntu-latest
    steps:
      - uses: langchain-ai/deepagents/action@main
        with:
          prompt: "Review and fix linting errors"
          model: "claude-3-5-sonnet"
          enable_memory: true
          timeout: 30

The action parameters:

prompt - What to tell the agent
model - Which LLM to use
enable_memory - Persist memory across runs
skills_repo - Load custom skills from GitHub
timeout - Maximum runtime in minutes

Remote Sandboxes

The CLI can run code in isolated environments instead of my local machine. Supported sandboxes:

┌─────────────┐     ┌─────────────┐
│  Local CLI  │────▶│   Sandbox   │
└─────────────┘     └─────────────┘
                          │
        ┌─────────────────┼─────────────────┐
        ▼                 ▼                 ▼
   ┌─────────┐      ┌─────────┐      ┌─────────┐
   │  Modal  │      │ Daytona │      │Runloop  │
   └─────────┘      └─────────┘      └─────────┘

This is useful when:

Running untrusted code
Testing in fresh environments
Isolating dependencies

Summary

In this post, I showed how to install and use Deep Agents CLI. It provides a terminal-based AI coding assistant with an interactive TUI, conversation persistence, web search, custom skills, and remote sandbox support. The one-line install makes it easy to get started, and the extensibility lets me customize it for my workflow.

Final Words + More Resources

My intention with this article was to help others share my knowledge and experience. If you want to contact me, you can contact by email: Email me

Here are also the most important links from this article along with some further resources that will help you in this scope:

👨‍💻 Deep Agents CLI Documentation
👨‍💻 Deep Agents GitHub

Oh, and if you found these resources useful, don’t forget to support me by starring the repo on GitHub!