How it works

Your AI doesn't reinvent the same fix twice. Before coding, it searches your brain. After solving, it contributes the lesson back. Two operations, one HTTP call each — and strict isolation so what you learn stays yours.

2 simple operations

Wire your agent to two endpoints and the loop runs itself. No SDK, no schema migration, no rebuild — just two POSTs the agent already knows how to make.

Op 1 — Search

Before writing a fix, the agent asks the brain. If a similar pattern has been solved before (by you or by the seed), it gets back the quick fix, root cause and origin instead of guessing from scratch.

Search operation: a new bug triggers a POST /api/v1/search call which returns matching patterns from your brain. New bug in your code POST /api/v1/search { q: "..." } Matching patterns quick_fix · root_cause Agent reads the result · skips re-inventing the wheel

If the brain has nothing useful, the agent falls through to its normal reasoning. No magic, no hallucinated answer — empty result means empty result.

Op 2 — Contribute

Once the agent has solved the new problem, it writes the lesson back. The pattern lands in your tenant brain — never in another tenant's, never automatically merged into the GNETICS master.

Contribute operation: after solving, the agent POSTs the new pattern which is appended to your tenant brain only. Agent solves new pattern POST /api/v1/contribute { title, fix, ... } Your brain tenant_id = you +1 pattern Stored in YOUR space — other tenants never see it

Promotion to the public master is opt-in and manual. We never lift your contributions silently into the seed.

Isolation diagram

Every row in the database carries a tenant_id. Every query — search, contribute, owner dashboard — is scoped by the key you authenticate with. The GNETICS master brain is read-only seed material: it never receives writes from tenants, and no tenant can read another tenant's data.

Two tenants each own a fully isolated brain. The GNETICS master brain is read-only and only seeds new tenants — tenants never write to it and never read each other. GNETICS master brain read-only · seeds new tenants seed seed Tenant A tenant_id = A Brain of A seed + A's own contributions Tenant B tenant_id = B Brain of B seed + B's own contributions

Your contributions stay yours. Other tenants never see them. Your AI never leaks across tenants. The isolated tenant model is enforced at the SQL layer — there is no code path where a query runs without the active tenant_id filter.

What's seeded into your brain

513 depersonalised patterns, pre-loaded on signup

The seed is derived from the real GNETICS master brain, curated and stripped of internal context (ticket IDs, repository paths, environment names, secret-shaped strings) before export. It's translated to English so it reads as a generic engineering corpus.

  • Runtime & process — Node, Express, pm2 (restart loops, env drift, log churn)
  • Persistence — SQLite (WAL, FTS, locks, migrations)
  • Edge & transport — Nginx (TLS, proxy_pass, headers), SSL renewal
  • Containers — Docker (compose ordering, volumes, network leaks)
  • Version control — Git (worktree, rebase pitfalls, stale refs)
  • General engineering lessons — debugging discipline, root-cause vs quick-fix, naming, retries

From day one the brain has 513 entries. From day two onwards, every problem your agent solves makes it bigger and more specific to your stack.

Plug your AI into a brain that learns

Free to start. One signup gets your tenant, your seed and your two keys (agent + owner).

Create my brain