Pontos OS
In Pontos OS, Kathras powers grounded product help that answers from the platform's own manual, along with a library of AI jobs the operator controls from Command.
Private AI, Made Clear.
Private AI for software that cannot afford to guess.
Kathras is the private AI engine behind Pontos OS and Tansy Social. It answers from approved sources, keeps every product behind its own door, records what it did without recording private content, and asks a person before anything important changes.
Kathras takes its name from the Greek katharos: pure, clean, clear.
Proof, not promises
Already working inside two real products
Tansy Social
In Tansy Social, the same core works through a completely separate door, running the AI review lane and the quiet Gardener upkeep jobs.
No training on your data
Nothing Kathras does trains on customer or member data. Each product's information stays behind that product's own door.
The engine
Named features, plain words
Ten capabilities, each one a named job with a switch, a log, and a person in charge. Two are live in production today; the rest belong to the engine's built job library.
Kathras Core
The shared engine at the center. One core serves every connected product, so a better engine lifts every product at once, while each product keeps its own data, rules, and voice.
Separate Doors
Every product connects through its own door, with its own permissions and data boundary. Nothing crosses between products, by construction rather than by policy.
Live in production
Grounded Answers
Help that answers from approved sources, such as the product's own manual, and names the source it used. When the source holds no answer, Kathras says so plainly. Live in production today in Pontos OS.
In the job library
Meaning Search
Search by meaning instead of keywords. A plain sentence finds the record or page that answers it, with tenant isolation applied before the model sees any data.
In the job library
Confirm Before Action
Kathras reads unstructured input, such as a forwarded confirmation or email, and prepares a clean structured record from it. A person reviews and confirms before anything is saved.
Job Control
Every capability is a named job with its own switch in the product's control center. Turn one off and the rest of that product, and every other product, carries on.
Clean Audit
Every job writes a log entry: which job ran, when, and what it decided. The entry never contains the private content the job worked on.
Graceful Fallback
If the engine stops, the product behaves exactly as it did before AI arrived. Each deployment passes an outage drill before the AI layer is trusted.
Live in production
The Quiet Gardener
Upkeep jobs that tidy, flag, and prepare work for human review, quietly and on schedule. Live in production today inside Tansy Social, beside its AI review lane.
In the job library
Watchman
A monitoring job that watches the products for failures and drift. It fixes only what is proven safe and reversible, and escalates everything else to a person.
Architecture
How it works
A short path, with the guardrails placed before the model, not after it.
- Products Pontos OS and Tansy Social. Kathras lives inside the products people already use, not beside them.
- Separate Doors One door per product, each with its own data boundary, permissions, and job switches.
- Kathras Core The shared engine. It routes each request to the right job and applies every guardrail first.
- Open Source Models The core runs on open source models, so the engine can operate privately, on hardware its owner controls.
Approved sources only
Kathras answers from sources the product owner has approved. If the answer is not there, it says so instead of guessing.
Isolation enforced in code
Tenant isolation is enforced in code before the model sees any data. It is never left to a prompt or a policy.
AI proposes, people dispose
Kathras prepares, drafts, and recommends. A person makes every consequential change.
Logged and switchable
Every job is logged without private content, and every job can be switched off from the product's own control center.
Ownership
Kathras is available for acquisition, licensing, and private deployment discussions.
Three routes, all held directly with Pontos Brands, the owner of Kathras. No prices are published; serious inquiries open a conversation.
Acquire Kathras
The engine, the Kathras brand, and full development rights. Pontos OS and Tansy Social retain the rights they need to keep running Kathras for their own users.
License Kathras
Integrate the engine into another platform under license, with the guardrails and job library intact. Pontos Brands keeps ownership of Kathras.
Private Deployment
A paid pilot, configured for one company's product and data boundaries, run privately before any wider commitment.
Ready to talk? Send an inquiry and it goes straight to Pontos Brands.
Inquire
Start a conversation
Tell us which route interests you. Submitting opens your email app with your note addressed to Pontos Brands; nothing is sent until you press send in your own mail program.
FAQ
Questions, answered plainly
Does Kathras train on customer data?
No. Kathras does not train on customer or member data. It reads only the approved sources a product gives it, and each product's data stays behind that product's own door.
Can it act on its own?
Only through named, allowlisted jobs that a product owner has switched on. Anything consequential needs a person to approve it first. Kathras proposes; people dispose.
What happens if the engine is unavailable?
The product keeps working exactly as it did without AI. Kathras is a layer inside the product, not a dependency, and every deployment passes an outage drill to prove it.
Is it a chatbot?
No. Kathras is an engine that lives inside products. It powers help, search, review, and upkeep jobs from within the software people already use. It is not a chat window bolted onto a page.