OnBuzz

Most agent orchestration tools require you to hand your workflow, your credentials, and often your data to a cloud service. OnBuzz inverts that: the vendor describes it as a local-first platform where you run a fleet of autonomous agents directly on your own machine, connecting to LLM providers yourself, with no cloud intermediary in the loop. The core workflow is agent definition, tool assignment, and task scheduling — all managed through an Electron-based desktop UI that ships with installers and binaries, not just a source repo you have to compile yourself.

The differentiating feature is the inter-agent collaboration model. Agents do not just run in isolation — the docs describe them working in parallel and handing tasks between each other, which means you can decompose a complex job (research, draft, review, publish) across a coordinated fleet rather than wiring a single monolithic prompt chain. This is the same architecture that powers Loxia’s commercial Autopilot product, so the open-source version is a production-derived design, not a stripped demo.

OnBuzz fits tightly scoped scenarios: local development of agent workflows before cloud deployment, privacy-sensitive workloads where data cannot leave the machine, and scheduled automation tasks where you need zero dependency on external uptime. It breaks down when you need a deep catalog of pre-built connectors, a managed scaling layer, or a large community to absorb your edge-case bug reports. The repository shows active development but a small contributor base, which means documentation gaps are real and workarounds often require reading source code.

The project is Apache-2.0 licensed and accepts community contributions via a public roadmap. It includes a skills directory, a web UI layer, and an Electron shell — the architecture is modular, but the integration surface for custom tooling is something you build and own, with precious little scaffolding from the community so far.