YYLO

YYLO coordinates coding agents from intent through validated code, keeping durable evidence of every task, session, review, and evaluation. Stop losing track of what your AI agents did and why—YYLO gives you full memory and accountability for AI-driven development.

• Full lifecycle coordination: from intent to merged code

• Durable session and task memory for context continuity

• Built-in review and evaluation evidence to ensure quality

• Works with any coding agents, providing audit-ready logs

Ideal for engineering teams shipping AI-generated code with confidence.

YYLO Ledger is part of the wider YYLO developer-tool ecosystem (https://yylo.dev/). Use it independently as a Git-based task management CLI, connect it with YYLO CLI (https://yylo.dev/projects/yylo) (GitHub (https://github.com/yylo-dev/yylo)) for AI coding-agent orchestration and guarded software delivery, or pair it with YYLO Benchmark (https://yylo.dev/projects/yylo-benchmark) (GitHub (https://github.com/yylo-dev/yylo-benchmark)) for evaluation and evidence workflows.

Product page: YYLO Ledger on yylo.dev (https://yylo.dev/projects/yylo-ledger)
Source code: YYLO Ledger on GitHub (https://github.com/yylo-dev/yylo-ledger)
Python package: yylo-ledger (https://pypi.org/project/yylo-ledger/) on PyPI (https://pypi.org/project/yylo-ledger/)
AI agent orchestration: YYLO CLI on yylo.dev (https://yylo.dev/projects/yylo) · GitHub (https://github.com/yylo-dev/yylo) · npm (https://www.npmjs.com/package/%40yylo%2Fcli)
Evaluation tooling: YYLO Benchmark on yylo.dev (https://yylo.dev/projects/yylo-benchmark) · GitHub (https://github.com/yylo-dev/yylo-benchmark) · npm (https://www.npmjs.com/package/%40yylo%2Fbenchmark)

Features

• Git-native task management: Store canonical task state in repository files that can be reviewed and versioned with the project.
• Developer-friendly command-line interface: Create, list, search, inspect, update, mark, tag, and archive tasks from a terminal or script.
• Dependency-aware planning: Define blockers, find ready work, detect invalid dependency transitions, and order tasks using calculated scores.
• Reviewable current state and history: Separate current Markdown state from segmented, append-only, hash-chained task history.
• Typed Record store: Manage task, wiki, workflow, artifact, and general Record profiles through an ID-first interface.
• Automation-ready output: Return bounded NDJSON, JSON, XML, or table output for people, shell scripts, coding agents, and CI systems.
• Safe concurrent updates: Use task locking, optional expected-revision checks, atomic current-state replacement, and bounded lock waits.
• Disposable query cache: Accelerate list and search operations with SQLite while keeping repository files—not the database—as canonical truth.
• Integrity and recovery tools: Run doctor, reconciliation, history checks, and cache rebuilds to verify or recover derived state.
• Immutable cold archives: Move eligible completed work into sealed archive packs while retaining exact retrieval and history access.
• Preservation-first migration: Inventory, plan, copy, verify, and receipt legacy wiki or artifact files without deleting the source material.
• Bounded and privacy-aware search: Control scope, projection, result count, and rendered bytes while keeping sensitive payloads out of default summaries.
• Read-only local Record host: Serve bounded Record projections to local tools without exposing mutation or workflow-execution APIs.
• Optional cross-project routing: Register explicitly allowed projects and query another initialized Ledger without unsafe fallback behavior.
• Shell completion: Generate completion for Bash, Zsh, and Fish.
• Open-source Python package: Install from PyPI (https://pypi.org/project/yylo-ledger/), inspect the source on GitHub (https://github.com/yylo-dev/yylo-ledger), and use it under the MIT License.