1 August 2026 — Lacunari is now available as open source under the GNU Affero General Public License v3.0. The project addresses a problem that has emerged sharply as AI coding agents became capable: the constraint on output is no longer how fast a model writes code, but that only a human knows what should be written next, and can only say so one prompt at a time.
Lacunari indexes a corpus — source code, scanned archives, scientific literature — into a structured map of what it contains, then derives what is absent from it: coverage holes in an otherwise continuous span, claims resting on a single source, entities connected to nothing, and material held but never read. Each absence becomes a task on a shared board that any participant can claim.
Coordination is handled entirely by Postgres. A claim is a single atomic UPDATE that also locks the files the task declared, so two workers cannot edit the same file — a refusal issued before work begins, rather than a merge conflict discovered after both parties have finished. A worker that crashes loses its lease, its files unlock, and its task returns to the board with its history intact.
Notable properties
- Offline by construction. Indexing, mapping and gap detection are set-based SQL. No model, no API key and no network call in that path.
- Vendor-neutral. Lacunari coordinates agents but does not run them. Claude, GPT, local Llama models, cron jobs and humans participate on equal terms.
- No daemon or broker. The entire substrate is a Postgres database.
- Enforcement in the database. Identity is the connected Postgres role, permissions are row-level security, and the audit log is append-only.
Licensing
Lacunari is licensed under AGPL-3.0. Commercial licences are available for organisations building it into proprietary products or offering it as a hosted service without publishing modifications.
About the founders
Evan King builds and operates the infrastructure Lacunari grew out of: a portfolio of production systems across print manufacturing, publishing automation and small-business operations, run with more machines than people. evanking.net
Matt Barrett has spent fifteen years in solutions engineering in the cybersecurity industry, beginning as employee #25 at a startup in 2007 and later at Cisco, and has led both pre-sales engineering teams and post-sales customer success. thesalesengineer.net