YYLO

Open-source system of work / agentic engineering

Run the work. Keep the memory.

YYLO (why-lo) coordinates coding agents from engineering intent through validated code, while preserving the task, session, review, and evaluation evidence that explains what happened.

Install YYLOView on GitHub
Open source · 0.2.1-rc.3npm ↗
~/your-projectbounded change
$ npm install -g @yylo/cli@0.2.1-rc.3
$ yy init --task
  "Add keyboard navigation" --subagent pi

intent explicit outcome
action bounded agent work
evidence tests · commit · response

✓ inspect what happened

System of work / available today

From intent to validated code—with evidence.

YYLO does not replace Git or your repository workflow. It coordinates agent execution and makes the work that produces code inspectable.

  1. 01Define intentA task records one bounded outcome, its constraints, and dependencies.
  2. 02Delegate executionYYLO routes work to an agent and preserves its worktree, model, and session provenance.
  3. 03Validate the resultTests, evidence receipts, and policy-bound review make acceptance inspectable.
  4. 04Record what happenedTask responses, commits, review evidence, and merge outcomes remain available for future work.

Available today

Control the work loop.

  • Manage task truth and dependencies in Git.
  • Delegate bounded work across agents and models.
  • Execute in isolated worktrees with session provenance.
  • Validate, review, and merge with durable receipts.
  • Compare agent attempts with retained benchmark evidence.

Direction

Build engineering memory.

YYLO is progressively connecting today's task and lifecycle evidence into an Engineering Change: the strategic record that links intent, decisions, attempts, validated code, and eventually real-world outcomes.

PR, deployment, production-outcome learning, and automatic workflow selection are product direction—not claims about the current release.

YYLO ecosystem / visual overview

One system. Three open-source layers.

The control plane coordinates action, the Ledger preserves task truth, and Benchmark turns attempts into evaluation evidence. The verified HTML below is authoritative; the infographic is a visual overview.

Visual overview of YYLO, YYLO Ledger, Git, and YYLO Benchmark as a connected agentic engineering workflow

One system / three layers

Coordinate, remember, evaluate.

Control plane · 0.2.1-rc.3

YYLO coordinates the work.

Delegate bounded agent execution, preserve provenance, validate candidates, and integrate accepted changes.

Explore YYLO →

Engineering memory · 0.2.1rc2

Ledger preserves task truth.

Keep intent, status, dependencies, responses, archives, and commit evidence durable in Git.

Explore Ledger →

Evaluation evidence · 0.1.0-rc.8

Benchmark compares attempts.

Plan immutable cases, isolate runs, reconcile artifacts, govern grading, and report with retained evidence.

Explore Benchmark →

Install / YYLO 0.2.1-rc.3

Start with one verifiable change.

Install the current release, choose a provider, and keep the first outcome small enough to validate and review.

Three-step quick start
npm install -g @yylo/cli@0.2.1-rc.3
yy init --task "Ship one verified change" --subagent pi
yy start -s pi -i 1

Begin with one change

Run the work without losing why it happened.

Install YYLOView on GitHub