Open-source project
ClawAI
A local-first AI orchestration platform that runs open models on your own machine and routes each prompt to the most suitable one.
Overview
Claw gives a team one workspace to manage, route, compare and operate many AI models at once. It prioritises private local inference through Ollama and llama.cpp, while still reaching cloud providers when a task needs them, so sensitive prompts can stay on hardware the organisation controls.
Architecture
A NestJS microservices backend sits behind a Next.js workspace, with PostgreSQL for relational data, MongoDB for documents, Redis for caching and queues for background work. A provider abstraction puts local and cloud models behind one interface, which is what makes routing, fallback and comparison possible without special-casing each vendor.
Engineering
The interesting problems were routing and trust: choosing a model per prompt, falling back when a provider degrades, and running comparison, consensus and judge workflows over several answers. Around that sit memory and context packs, workspace grounding, file processing, permission-aware tool execution for connected business tools, provider credential handling and health aggregation.
- Role
- Architected and built the platform end to end — microservices backend, orchestration layer and web workspace.
- Category
- AI
- Stack
- Next.jsNestJSTypeScriptMicroservicesPostgreSQLMongoDBRedisOllamallama.cppDocker
- Language
- TypeScript
- Licence
- Apache-2.0