Open source

Recall is open source.

The Rust core, the TypeScript and Python SDKs, the Learn-track interactive demos, and the spec are all open. We build in the open because cognitive infrastructure should be inspectable.

License

Apache 2.0 for the Recall core, SDKs, and Learn-track demo components. Use it commercially, modify it, ship it inside your product — permissively. No CLA required for contributions.

Some auxiliary tooling (benchmarks, internal evaluation harnesses) lives under separate repositories with the same license.

What's in the open

  • recall (core)
    Rust core: pipelines, retrievers, indexes
  • @arc-labs/recall (TS)
    TypeScript SDK via napi-rs
  • recall (Python)
    Python SDK via PyO3
  • Learn demos (apps/arc-labs)
    D3 + SVG demos used on this site
  • spec/
    The architectural spec — write/read/math/drift
  • evals/
    Reproducible eval harness (in progress)

All under github.com/arc-labs.

Governance

Today: project governance is held by the Arc Labs core team. Major architectural decisions are documented as spec proposals in the spec repository, open for review before merge.

We anticipate moving to a more formal governance model (steering committee + RFC process) once contributor counts justify it. For now, opening an issue on GitHub or emailing opensource@arc-labs.ai is the path.

Contributing

The contribution path:

  1. Open an issue describing what you want to do (feature, fix, doc).
  2. Wait for maintainer feedback before significant work — saves both sides.
  3. PRs require: tests, docs if user-visible, no new dependencies without discussion, all CI checks green.
  4. We review within 5 business days. Honest feedback; small PRs preferred.

We don't require a CLA. Contributions are licensed under the project's Apache 2.0 by submission.

Security policy

Vulnerabilities go to security@arc-labs.ai, not GitHub issues. Acknowledged within 24 hours; critical fixes within 14 days. We credit reporters with permission. See /security for the full policy.

Roadmap

The roadmap lives in the spec repository as /research. We update it quarterly. Major near-term items:

  • Reproducible benchmark harness (Q3 2026)
  • Plan primitive — initial preview (Q3 2026)
  • Multi-region replication for managed cloud (Q4 2026)
  • Reason primitive — initial preview (Q1 2027)

Get involved

Star the repo, open an issue, send a PR, or write to us.

Updates from the lab.

Engineering notes, research drops, occasional product updates. Roughly monthly.