Free public-repo trial · Open source

Is your repo ready for an agent team?

One agent can improvise. A team needs shared rules. Scan the repository contracts that keep parallel coding agents aligned, isolated, and verifiable.

Static and read-only. Your code is inspected, never executed.

6team-readiness dimensions
atr-1machine-readable evidence
0repository writes
Example Agent Team Readiness report for First Tree, showing a score of 84 with 99 percent evidence coverage
Actual scanner output First Tree · 2,079 files analyzed
See what it measures
Instructions Isolation Ownership Context Verification Handoff

The six-dimensional rubric

Measure the repo your agents actually inherit.

The scan grades observable repository support for multi-agent work. Every score traces back to bounded evidence, and uncertainty stays visible instead of being converted into confidence.

  1. 01 18%

    Instruction convergence

    One discoverable root contract, with scoped guidance that agrees instead of sending agents in opposite directions.

  2. 02 18%

    Task & workspace isolation

    Branch, worktree, and shared-state rules that let parallel tasks move without quietly overwriting each other.

  3. 03 16%

    Code & domain ownership

    Review paths and ownership boundaries clear enough for an agent to route decisions instead of guessing.

  4. 04 16%

    Shared decisions & context

    Durable product and architecture context that survives beyond one prompt, one chat, or one agent run.

  5. 05 20%

    Repeatable verification

    Real, runnable test, lint, build, and CI paths — not empty scripts, placeholders, or prose that only looks executable.

  6. 06 12%

    Handoff & definition of done

    A shared finish line: what to report, what evidence to attach, and how the next human or agent picks up the work.

Built for inspection

A score is only useful when you can challenge it.

The hosted report separates facts, inferences, and unknowns. Drill into the exact file evidence, then carry the machine-readable report into a repair workflow without scraping a dashboard.

  • 01
    Evidence coverageKnow how much of the repository was actually analyzed before trusting the headline score.
  • 02
    Prioritized blockersFixes are ranked deterministically by severity, dimension weight, and distance from a strong contract.
  • 03
    Review-only artifactsDraft instructions and a Context Tree seed map are outputs to review — never silent repository edits.
Inspect the open-source scanner
atr-1.json validated 
{
  "schema_version": "atr-1",
  "headline_score": 84,
  "evidence_coverage": 99,
  "dimensions": [
    // score · evidence · unknowns
  ],
  "artifacts": [
    "AGENTS.draft.md",
    "context-tree-seed-map.md"
  ]
}
Deterministic Schema-validated Portable

Read the score narrowly

Repository readiness, not a claim about your people.

The scan cannot certify runtime isolation, permissions, owner availability, or actual team behavior from repository files alone. Those stay explicit unknowns. That boundary is part of the product, not a footnote.

From URL to evidence

Three steps. No repository access ceremony.

  1. 1

    Paste a public repository

    Give the scanner a GitHub URL. No install, token, or write access is requested for the public-repo trial.

  2. 2

    We inspect — never execute

    The scan inventories repository evidence deterministically. It does not run project code or trust commands found inside it.

  3. 3

    Get evidence, not vibes

    Receive an atr-1 report, cited blockers, and review-only AGENTS.md and Context Tree starter artifacts you can inspect before adopting.

Your next agent should not start from zero.

Give the whole team a better starting point.

Scan a public GitHub repository and see the exact contracts that help — or hinder — parallel agent work.

Public repositories only · Read-only static analysis · Limited daily capacity