Design AI agent workflows visually and run them from the Windows tray
Experience a canvas-first way to assemble AI agent teams: drag blocks on a visual canvas that run as executable logic, with block-level logs and re-run controls. Concentus, by Gerard Gilabert, runs locally using your existing browser sessions instead of requiring API keys or servers, and sits in the Windows system tray. It suits AI enthusiasts and developers who want to build multi-agent automations without managing infrastructure, offering undo/redo history and per-block debugging for iterative work.
What does Concentus do and how does it execute flows?
Concentus builds AI-driven workflows on a visual canvas where each connection is the actual runtime logic, not a separate config file. The app orchestrates multiple agents in one flow and runs tasks using local browser sessions (for example, an existing Claude login) rather than requiring API key setup or remote servers. The canvas supports alignment guides and snap-to-grid placement so flows remain readable as they grow.
Does it affect system performance during operation?
Concentus runs as a lightweight Windows utility from the system tray, designed to avoid heavy background load. The tool executes via the desktop environment and local browser context, so CPU and RAM impact depend on the browser sessions it drives rather than a persistent server process. Installation is available as a standard executable or via Winget for scriptable deployment on Windows machines.
Is it safe to use in a local environment?
Concentus uses local-first execution and does not require centralized servers, which reduces the need to share API keys or external databases. Block-level debugging records input and output for each step and lets you re-run individual blocks, so you can test changes without replaying entire flows. The project is hosted on GitHub, which provides project transparency and source access.
Does it require specialist technical knowledge to operate correctly?
The canvas-first approach lowers the barrier compared with manual API wiring, but users should understand browser authentication and agent choreography to avoid misconfigurations. Undo/redo and granular re-runs help iterative users, while agent-team management and token tracking make the tool useful for developer workflows that need visibility into token usage and run estimates.
Concentus is practical for builders who want local, visual AI automation
The tool's visual-to-logic parity and local browser execution make it a pragmatic choice for people who need to prototype multi-agent automations without API administration. It provides clear debugging and history controls for iterative refinement. One trade-off is dependence on local browser sessions for authentication, which ties execution to the desktop environment rather than a headless server.
