SuperAPI AI Directory
Partner with us中文AI newsFavorites
DirectoryNews and tutorials

yoyo - An open-source self-evolving AI coding agent

Author: AI小集 · Original source: AI工具集

What is yoyo

yoyo is a self-evolving AI codingagent living in cyberspace. After developer Yuanhao started it with 200 lines of Rust code, it was fully handed over to AI for autonomous iteration, growing from zero human commits to 47,000+ lines of code within 47 days. The tool automatically wakes up every 8 hours, reads its own source code, analyzes GitHub Issues, plans improvements, runs tests, and commits code; every 4 hours it conducts social conversations to learn from community feedback.

yoyo

yoyo's main features

  • Autonomous evolution loop: Periodically reads its own source code and community Issues, autonomously plans features, writes code, runs 1,700+ tests, and commits, forming a closed-loop iteration.
  • Multi-file code editing: Supports cross-file navigation, precise text replacement, symbol renaming, and multi-file batch modifications.
  • Full Git workflow integration: Built-in 60+ slash commands for diff, blame, commit, PR creation and review, branch management, and more.
  • Multi-model support: Compatible with 12 providers including Anthropic, OpenAI, Google, DeepSeek, and Groq, and can be switched at any time during a session.
  • Sub-agent delegation: Via /spawn or automatic tools to dispatch subtasks to sub-agents for parallel processing.
  • Project health diagnostics: Automatically detects project types such as Rust, Node, Python, and Go and runs builds, tests, linting, and fixes.
  • Memory and context: Automatically loads YOYO.md project context, supports persistent memory across sessions and daily learning archive compression.
  • Community interaction: As a bot, automatically replies to Issues and Discussions, prioritizing community requests based on upvote counts.

yoyo's technical principles

  • Harness constraint framework: Encapsulates software engineering standards as Agent operating rules, enabling AI to make autonomous decisions within restricted but safe boundaries, reducing the risk of losing control.
  • Dual-loop architecture: The evolution loop (code self-improvement) and the social loop (community interaction learning) run in parallel, using time-weighted compression to archive historical learning as active memory injected into prompts.
  • Tool calling and streaming output: Implements streaming Token output and parallel tool execution based on Rust, supporting atomic operations such as Bash, file reading and writing, code search, and AST structural queries.
  • Provider Failover: Automatically degrades and switches to backup models by priority when API failures occur, ensuring stability during long autonomous runs.
  • Mutation testing: Uses cargo-mutants to perform mutation testing on the test suite, ensuring the true effectiveness of code coverage.

How to use yoyo

  • Installation: On macOS/Linux run curl -fsSL .../install.sh | bash, on Windows use the PowerShell script, or cargo install build from source.
  • Configure API Key: After setting ANTHROPIC_API_KEY or another provider Key, enter yoyo in the terminal to start the interactive REPL.
  • Project initialization: Run /init in the project root directory to automatically generate the YOYO.md context file, or create it manually to inject domain knowledge.
  • Daily use: Via natural language or 60+ slash commands (such as /diff、/commit、/test、/spawn) drives coding tasks.
  • Community participation: Submit Issues on GitHub with the agent-input label and upvote them to influence yoyo's evolution priorities; or start a Discussion to converse with it.

yoyo's key information and usage requirements

  • Open source license: MIT, free to fork and cultivate your own evolving agent.
  • Runtime environment: Rust toolchain, supports macOS, Linux, Windows.
  • Model dependencies: You need to bring your own API Key (Claude Sonnet 4 recommended), supports 12 providers compatible with the OpenAI interface.
  • Permission control: Interactive confirmation for sensitive operations by default, supports --yes global auto-approval or glob-based allowlist/denylist fine-grained control.
  • Sponsorship mechanism: GitHub Sponsors / Ko-fi monthly or one-time sponsorship can boost Issue priority and accelerate run counts.

yoyo's core advantages

  • Zero human code: All code is autonomously written and submitted by AI, the evolution process is fully open and transparent, refusing to reference leaked source code to maintain independent growth.
  • Build in Public: All decisions, diaries, learning records, and code changes are viewable in real time, forming a reproducible Agent growth path.
  • Strong engineering constraints: The Harness mechanism internalizes software engineering best practices as the Agent's behavioral boundaries, balancing autonomy and code quality.
  • Community immunity mechanism: The Issue voting system lets the community jointly safeguard the evolution direction, suppressing prompt injection and malicious interference.

yoyo's project address

  • Project website：https://yoyo.yolog.dev/
  • GitHub repository：https://github.com/yologdev/yoyo-evolve

yoyo's comparison with similar competitors

Dimension yoyo Claude Code OpenAI Codex CLI
Code author AI autonomously written, zero human commits Developed by a human team Developed by a human team
Evolution mechanism Autonomous loop: read source code → plan → implement → test → commit Depends on official version updates Depends on official version updates
Degree of open source Fully open source (MIT), can fork and self-evolve Closed source Closed source
Community interaction Automatically replies to Issue/Discussion, voting priority None None
Operating mode Scheduled wake-up for autonomous operation + interactive REPL Pure interactive CLI Pure interactive CLI
Memory system Daily synthesized compressed learning archive injected into prompts Session-level context Session-level context
Model support 12 providers can be switched dynamically Anthropic only OpenAI only
Target positioning An open and transparent self-evolving lifeform experiment Enterprise-grade coding assistant Enterprise-grade coding assistant

yoyo's application scenarios

  • Automated maintenance of open source projects: As a 24/7 non-stop collaborator, automatically refactors code, fixes bugs, manages GitHub Issues, and replies to community feedback.
  • Terminal intelligent programming assistant: Developers invoke 60+ slash commands through a streaming REPL to complete codebase understanding, multi-file editing, test execution, and the full Git workflow.
  • AI Agent evolution research: After researchers fork the project, they modify IDENTITY.md and PERSONALITY.md, set different goals and personalities, and observe the Agent's autonomous growth trajectory in a public environment.
  • Programming education and decision-making learning: By reading yoyo's diary, commit history, and learning archive, observe how AI derives software engineering decisions and code refactoring strategies from first principles.
  • Community-driven product iteration: Product teams submit requirements in the form of GitHub Issues, priorities are determined by community voting, and yoyo automatically implements high-vote features and continuously delivers.