deepagents-code

What it is and what it does

deepagents-code is a pre-built terminal agent that lets you interact with any LLM that supports tool calling to write, debug, and run code—similar to Claude Code or Cursor but in your terminal. It wraps the Deep Agents SDK with a rich interactive TUI, conversation persistence, web search grounding, and optional sandboxed execution. You install it with a single command and run `dcode` to start; no Python code required from you.

The agent operates in your working directory by default, reading project files to inform its responses and executing code with human-in-the-loop approval gates. It supports 35 runtime dependencies including langchain, langgraph, pydantic, and model-provider integrations (langchain-openai, langchain-anthropic, langchain-google-genai). For untrusted codebases, you can route execution to remote sandboxes to isolate it from your machine. The package is actively maintained, supports Python 3.11–3.14, and carries an MIT license.

Use it for:

• Rapid prototyping and debugging: ask the agent to write, test, and fix code interactively in your terminal without leaving the shell.
• Code review and refactoring: feed the agent your project files and ask it to suggest improvements or refactor sections with approval before changes.
• Learning and experimentation: use it as a hands-on tutor to explore new libraries or algorithms with live code execution and feedback.
• CI/CD scripting: run dcode in headless mode to generate or validate code as part of automated pipelines.
• Untrusted repository inspection: analyze third-party code in a remote sandbox to prevent malicious execution on your local machine.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

A terminal-based AI coding agent that works with any LLM supporting tool calling, providing an interactive interface for code generation, execution, and debugging without requiring code setup.

Yes. Active maintenance, low install friction, no known vulnerabilities, permissive MIT license, and a large runtime dependency graph that is well-established (langchain, langgraph, pydantic). Best for developers who want an interactive LLM-powered coding assistant in the terminal and are willing to configure an LLM provider. Avoid running it on untrusted code without a remote sandbox backend.

Install

deepagents-code on PyPI

pip install deepagents-code

uv add deepagents-code

poetry add deepagents-code

Installing deepagents-code

Before you install

Low friction: pure Python wheel with no compiled dependencies. Active maintenance with a release on 2026-08-14 and 27776 repository stars. Requires Python 3.11 or later.

License in practice

MIT license permits unrestricted use, modification, and distribution for commercial and private purposes with minimal attribution requirements.

Quickstart

curl -LsSf https://langch.in/dcode | bash
dcode

Requires Python 3.11 or later and a configured LLM provider (OpenAI, Anthropic, or Gemini included by default; others available via extras).

Verify before relying

• Whether remote sandbox backends (LangSmith, AgentCore, Daytona, Modal, Runloop) require separate credentials or setup beyond the base install.
• Performance characteristics when handling large codebases or long conversation histories.
• Whether custom skills can be added without modifying the package source or if there is a plugin system.

Package facts

Evidence: deepagents_code-0.1.56-py3-none-any.whl

Keywords: agents, ai, cli, deep-agent, langchain, langgraph, llm, terminal

Tags

More Python Modules packages

Further reading