CLAT turns open-source agent innovations into a compact, trilingual AI assistant that runs entirely on your own machine — no cloud, no heavy setup, any AI model.
Not a browser tab talking to someone else's server — a full agent platform: terminal, web and desktop clients, running where your files already are.
Everything runs locally. Code, documents and history stay on your machine — safe for finance, healthcare and government work.
Switch between DeepSeek, GLM and any OpenAI-compatible service at will. No vendor lock-in, ever.
One small program, minutes to install, runs on ordinary laptops — macOS, Windows and Linux.
The assistant asks before it acts. Durable tasks, verified results, nothing silently executed.
Open-source communities such as DSH invent frontier agent capabilities every week. Big tech pushes intelligence forward. Almost nobody makes these results small, simple and cheap to run. That is CLAT's job.
Projects like DSH create cutting-edge agent plugins, tools and session formats — powerful, but heavy and expert-oriented.
One small binary, exact ecosystem compatibility, trilingual interfaces. Frontier capabilities without frontier complexity.
SMEs, schools, public bodies and developers across Hong Kong, Shenzhen and the Greater Bay Area — served on machines they already own.
CLAT v2.2 ships now, hardened by 2,400+ automated tests and used daily by its own builders.
Near-zero deployment cost lets CLAT serve more people with limited resources — and gives ARTEC software-grade margins.
Start with developers and engineering teams — the fastest adopters — then expand to SMEs, schools and public bodies across the GBA, plus global users of lightweight tools.
A free open-source core builds a global user base at near-zero marketing cost. Revenue from subscriptions, enterprise licences, support and an integration marketplace.
Giants cannot copy "small and local" without breaking their cloud business. Ecosystem compatibility brings innovations to CLAT almost for free; users accumulate tasks, sessions and plugins.
v2.2 live across terminal, web and desktop. Durable tasks, verified completion, permission-gated execution — the hardest requirements, already met.
Finance and technology firms deploy CLAT locally: private, compliant, running on hardware they already own. One lightweight host serves multiple assistants.
The same engine extends to documents, data and daily workflows — a ready-to-use integration marketplace brings assistants to every knowledge worker.