Velocity

Why we included this project

Velocity is a web-native, agent-first IDE that treats the model as a pluggable component rather than a fixed dependency. Instead of being tied to one vendor's assistant, you route work through a universal LLM gateway and use whatever model fits the task, from frontier models to cheaper local ones. The backend already handles the messy parts of multi-agent coordination: a SQLite layer with sequential migrations, hierarchical agent swarms, a shared message bus, file-level locks to avoid edit collisions, and memory snapshots for recovery. It also tracks token usage and per-call cost with configurable spending caps, which matters when you let agents run long tasks. Teams that want to self-host and keep code and data on their own infrastructure will appreciate the container-native setup and privacy-first stance, though the project is still experimental and best treated as a foundation to build on rather than a drop-in production tool.