Weft
Self-hosted AI workflow language with a Docker-based runtime and VS Code graph editor. Its WhatsApp bot example calls models through OpenRouter.
Weft is a language and framework for developers who want to run AI workflow orchestration on their own machine. Its runtime requires Docker, and its graph editor works in VS Code and compatible forks. Self-hosting the workflow doesn't make its model calls local: the included WhatsApp support bot uses OpenRouter for inference.
Workflows describe how models, APIs, databases and human decisions connect, with each node receiving the context assigned to its task. The compiler checks connection types, required connections and graph structure before building a native Rust program. Nodes can remain active while other work continues, and a workflow can save a wait for a person or external event and resume when a response arrives.
The graph editor displays execution records and the values passed between nodes. Recording can be disabled for a workflow's trigger. Human approvals use the Weft tasks browser extension; the WhatsApp example holds sensitive replies for review before sending them.
Tangle works with coding assistants such as Claude Code, Codex and Cursor to build workflows, test individual nodes and compare later changes against accepted examples. Using Claude Code sends prompts and model context over the network to its provider. Tangle can also build a web interface alongside a workflow.
Weft uses the O'Saasy License. Developers can write custom nodes in Rust, and the runtime manages infrastructure containers such as Postgres with storage that survives restarts.
