Labs

Built by us, for us.
Proven before they’re products.

Labs

Built by us, for us.
Proven before they’re products.

Labs is where the technology we build for our own work becomes products in their own right. Building SeKondBrain means building serious tools along the way — not demos, but systems we depend on daily. Labs is where they graduate: proven inside one Brain, extracted into standalone services, hardened, and offered to everyone — priced by usage, in credits.

How Labs works

Proven inside. Extracted.
Offered to all.

Every Labs initiative follows the same path. It’s born solving our own problem inside a Brain. When it proves itself, we extract it into an independently versioned, separately deployable service. From there it takes one or both of two routes:

  1. Step 1Born inside a BrainSolving our own problem
  2. Step 2Extracted into a standalone serviceIndependently deployable
  3. Step 3A credit-priced productAPI-first, metered by usage
  4. Step 4An open-source community releaseLives on Community

Some initiatives take one route, some both. All of them keep improving inside our products while they live their own lives outside them.

How Labs works

Born inside a Brain.
Never staying there.

Every initiative starts the same way: born solving a problem inside one of our own Brains. Once it proves itself, we extract it into an independently versioned service — kept open-source, offered as a credit-priced product, or both, while it keeps improving inside the products that spawned it.

The initiatives

What’s on
the bench.

Live in staging · credit-priced
RedaKtion
Try RedaKtion

Not a black box. A clean burn. Finds every piece of sensitive information in a document and permanently removes it — native PDFs, scans and phone photos, office files — across English, Arabic and Hindi. Most redaction draws a rectangle over text that’s still there; RedaKtion deletes the characters, overwrites the pixels, strips the metadata, then re-reads its own output and re-runs the whole detection stack against it. If risk remains, it says so — a broken redaction never ships under a green tick.

In development
Code Graph
Register interest

Your codebase as a concept graph — repositories, features, components and their relationships as connected, queryable structure rather than files in folders. The layer that lets agents and people reason about what code means and what a change touches, not just what it says.

In beta
DevOps Agent (Adam)
Register interest

Deployment as understanding, not scripts. Adam holds environments, services and deployments as first-class concepts in the graph — builds and maintains CI/CD, chooses deployment patterns by risk, enforces quality gates, and writes every runbook and operational decision back into the brain, so ops knowledge compounds instead of living in one engineer’s head.

45 configurable categories6 detection layers0 silent failures9 patents pending

RedaKtion plans: 500 pages free, every feature on — then pay as you go by the page.

detection stack · pass 1 · detect
scanning 45 categories…
RedaKtion

Not a black box.
A clean burn.

Most redaction draws a rectangle over text that’s still there. RedaKtion deletes the characters, overwrites the pixels, strips the metadata — then re-reads its own output and reruns the whole detection stack against it, so a broken redaction never ships under a green tick.

Credits

Buy the work,
not the seat.

Every Labs product is metered the same way: credits that map to compute. Verification is always included, capability is never gated, and free tiers are real — enough to prove it on your own material, no card, no watermark.

Labs and Community

Two shelves,
one workshop.

Labs is where initiatives are productised and priced. Community is where their open-source releases live — packaged versions you install and run yourself. Kemory’s Community Edition is there today; Labs initiatives join it as we package open-source versions.

Code Graph

Code as structure,
not files in folders.

Repositories, features, components and their relationships become connected, queryable structure instead of files in folders — the layer that lets agents and people reason about what code means and what a change touches.

What’s coming

The bench
keeps filling.

Project Hai · in active iteration · repo: Kora_Hai

Agents that heal themselves — grown from a seed.

Hai (胚) is Japanese for embryo, and that’s the design: an embryonic kernel, born with a handful of seed skills, that grows into whatever agent the task demands. When an agent meets a task it can’t do, Hai checks the shared brain first — and if the skill doesn’t exist, it writes it.

More initiatives are extracted as they prove themselves inside the Brains. Pick one, put it to work, and tell us how it goes — real use is what turns an initiative into a product.

Labs

Pick one.
Put it to work.

Real use is what turns an initiative into a product. Explore what’s on the bench, or run the open-source releases in Community.