#7777th of 142 in Agent frameworks and SDKs
kungfu
Durable Work layer that keeps tasks intact across coding agents and failures
Documentation ↗
(opens in a new tab)Website ↗
(opens in a new tab)Repository on GitHub ↗
(opens in a new tab)
- Stars
- 4.5k
- License
- Apache-2.0
- Last commit
- Sep 2026
- Last release
- Sep 2026
- Language
- C++
3,814 on 21 Apr 20264,541 on 10 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Kungfu is a C++ CLI, TUI and GUI that stores a task (Work) outside any agent session, with its objective, progress, evidence and next action. It wraps agents such as Codex, Claude, OpenCode and Amp via `kungfu run <agent>`, records each Attempt including failures, and blocks a second live writer on the same Work. Completion is decided by independent review and settlement, not by the agent that did the work. Alpha.
Who it is for: Developers who switch between coding agents and want durable task state
Strengths
- Work state persists across agent switches, crashes and disconnects without copying chat context
- Blocks a second live agent from writing to the same Work
- Built-in deterministic Mock Agent needs no provider credentials to try recovery scenarios
- Agent cannot mark its own work complete; independent review and settlement decide
Weaknesses
- Alpha prerelease; README says it is not stable or generally available
- Regular onboarding still needs a supported real agent for independent review
- Creates a project-local .kungfu/ directory with Git-staging rules you must follow
- Heavy custom vocabulary (Project, Work, Attempt, KFD, Shifu) raises the learning curve
What it needs
- no GPU
- Needs Codex, Claude, OpenCode or Amp (for real agent runs and review)
- Models: Codex, Claude, OpenCode, Amp
Also in Agent frameworks and SDKs
See all 142| Rank | Project | Score |
|---|---|---|
| 1 | ppt-masterAgent skill that turns documents into natively editable PowerPoint decks | 96 out of 100 |
| 2 | openai-agents-pythonPython SDK for multi-agent workflows with handoffs, guardrails and tracing | 92 out of 100 |
| 3 | mastraTypeScript framework for building AI agents, workflows and MCP servers | 92 out of 100 |
| #4 | browser-usePython library that lets an LLM agent control a web browser | 90 out of 100 |
| #5 | copilotkitReact and Angular SDK for adding agent chat and generative UI to apps | 90 out of 100 |