#173173rd of 236 in Coding agents and AI editors

pydantic-deepagents

Terminal coding assistant and Python agent framework on Pydantic AI

Stars
1.1k
License
MIT
Last commit
Oct 2026
Last release
Oct 2026
Language
Python
Stars, last 29 weeks+421
25 Apr 2026: 700 stars2 May 2026: 725 stars9 May 2026: 757 stars16 May 2026: 776 stars23 May 2026: 799 stars30 May 2026: 816 stars6 Jun 2026: 836 stars13 Jun 2026: 854 stars20 Jun 2026: 874 stars27 Jun 2026: 890 stars4 Jul 2026: 936 stars11 Jul 2026: 956 stars18 Jul 2026: 967 stars25 Jul 2026: 991 stars1 Aug 2026: 1,005 stars8 Aug 2026: 1,013 stars15 Aug 2026: 1,025 stars22 Aug 2026: 1,034 stars29 Aug 2026: 1,046 stars5 Sept 2026: 1,054 stars12 Sept 2026: 1,058 stars19 Sept 2026: 1,065 stars26 Sept 2026: 1,067 stars3 Oct 2026: 1,075 stars10 Oct 2026: 1,078 stars17 Oct 2026: 1,102 stars24 Oct 2026: 1,107 stars31 Oct 2026: 1,112 stars7 Nov 2026: 1,116 stars14 Nov 2026: 1,121 stars

700 on 25 Apr 20261,121 on 14 Nov 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

A Claude Code-style terminal assistant that works with any tool-calling model, plus the Python framework behind it. create_deep_agent() gives a model a filesystem, shell, planning, memory, sub-agents, MCP client and an optional Docker sandbox. It also offers Live Run Forking, which splits a run into isolated branches and merges the winner by AI judge, test results or vote.

Who it is for: Developers wanting a model-agnostic Claude Code alternative or an agent library

Its own README calls it an alternative to Claude Code.

Ranked among open-source alternatives to Claude Code (#10 of 10).

Strengths

  • Same harness works as a TUI (pydantic-deep) and as a Python library
  • Supports Anthropic, OpenAI, OpenRouter, Gemini, Ollama or any OpenAI-compatible endpoint
  • Run forking with copy-on-write overlays, per-branch budgets and a test-command merge hook
  • Headless `pydantic-deep run` for CI, plus MCP client, hooks and USD budget limits

Weaknesses

  • Comparison table and uniqueness claims are self-reported by the authors
  • Many capabilities are opt-in flags that need configuration
  • Browser automation needs the separate pydantic-deep[browser] extra
  • Docker sandbox needs a local Docker install; not bundled

What it needs

  • no GPU
  • Needs Python 3.10+, Pydantic AI, LLM provider API key or Ollama, Docker (optional, sandbox), Playwright (optional, browser)
  • Models: Anthropic, OpenAI, OpenRouter, Google Gemini, Ollama
  • README: alternative to Claude Code

What changed

From the release notes. No flag means the note does not say so, not that the upgrade is safe.

Also in Coding agents and AI editors

See all 236
Also in Coding agents and AI editors
RankProjectScore
1claw-codeRust CLI coding agent harness using Anthropic and OpenAI-compatible APIs194.8k starsMIT96 out of 100
2t3codeDesktop, web and mobile front end for controlling local coding agent CLIs Live demo ↗ (opens in a new tab)64.4k stars+36,149 (+128%) in 30 daysMIT96 out of 100
3codewhaleTerminal coding agent in Rust that works with 40+ model providers41.2k starsMIT94 out of 100
#4gooseRust AI agent with desktop app, CLI and API, extended through MCP56.7k stars+1,476 (+2.7%) in 30 daysApache-2.093 out of 100
#5ArchonYAML workflow engine that runs AI coding agents through fixed dev steps23.9k starsMIT93 out of 100