Langbase is a serverless platform for building and deploying AI agents and AI features from composable primitives rather than self-managed infrastructure. The central unit is a Pipe: a configured agent endpoint that an application runs by name. Memory covers retrieval over uploaded documents and repositories, and tools let a pipe fetch external data during a run.
Applications reach the platform through SDKs in TypeScript and Python. The TypeScript client is constructed with an API key; running a pipe takes its name and a list of messages and returns either a complete response or a stream that a runner consumes as content events. Worked examples cover Node.js scripts and a Next.js application with route handlers and a React component that displays the response.
Memory presents retrieval-augmented generation as an API, bundling a vector store, file storage, and a retrieval engine, with agentic re-ranking of retrieved context and citations back to the source. The platform also offers Command Code, a coding agent built on a model called taste-1 that adapts to a codebase's conventions from explicit and implicit feedback, with those preferences pushed and pulled so a team shares them.
Features
- Pipes: named agent endpoints run through the SDK, returning a complete response or a stream
- Memory: managed retrieval with vector store, file storage, and agentic re-ranking with citations
- Tools: a pipe calls tools mid-run to look up external data
- SDKs: TypeScript and Python clients authenticated with an API key
- Streaming: a runner emits content events as the response arrives
- Framework examples: Node.js scripts and a Next.js app with API route handlers and a React display component
- Command Code: a coding agent whose taste-1 model learns a codebase's conventions and shares them across a team
