Mentant

Secure, dependable, open source agents.
No fluff. No compromises.

Mentant
/ˈmɛn.tənt/
n. A reliable, thinking AI agent designed to reason persistently, maintain context across extended tasks, and execute dependably without supervision, cloud lock-in, or unnecessary complexity.
Etymology: Blend of Latin mēns, mentis (“mind; rational thought; intention; understanding”) + English agent (from Latin agēns, present participle of agere “to do, to act, to drive”).
Latin mēns (root men- “to think”) is the source of English “mental”, “mention”, dementia (dē- “away” + mēns), and the suffix -ment (denoting result, means, or agent of action).
Post-OpenClaw discipline

Built for business and users that expect the thing to work, not require babysitting.

Grok First Design

Utilising the latest grok models for orchestration, cheap, fast, truth seeking.

Open Source, local-first

Runs on your hardware. No cloud vendor lock-in. Your data stays yours.

Memory that... remembers

Persistent memory with full-text & semantic search. No more forgetting on reset.

Mentant Context Tree

Each agent has a structured way of organizing and accessing its knowledge.

The Death of Heartbeat.md

You don't need a model to deterministically decide if something should be run, that's what databases are for.

Rails-native extensibility

Providers, Channels & Tools are plain Ruby gems. Write one, bundle install, it’s live, no 500k LOC config hell.

A dashboard you can use

Simple, clean and elegant with what you need in front of you and available when you need it.

Scales when you need it

Starts lean on SQLite with Vector search. Needs MOAR power? Switch to Postgres with one config change.

Launching soon.
Be the first to know.

No spam. One email when we launch.