Hey everyone! I'm the developer behind Semwright. I started working on this because I kept running into the same problem when trying to get AI agents to work with desktop applications. Getting one operation working was usually manageable. Getting several applications to work together, keeping track of their state, and making the whole thing reusable was where things got complicated. And I kept thinking, why should every agent have to build its own way of interacting with the same software? That's what led me to Semwright. I wanted a common runtime where applications expose their capabilities through structured drivers, and different agents can use them without rebuilding the integration every time. One thing I've spent a lot of time on is what happens after an agent performs an action. Knowing which objects were modified, how different resources depend on each other, and whether the expected result actually happened is just as important as getting the operation to run. I've just released v1 as open source. It includes integrations across 3D, design, documents, browsers, video, and audio, along with SDKs for building additional integrations. There's also a short demo showing a workflow across Blender, Motion Canvas, and MLT, but that's just one example of what the runtime can be used for. I'm curious what people here are trying to automate outside of coding. Are there applications or workflows you'd want an agent to work with but haven't found a good way to connect yet?

Semwright

Give AI agents structured access to real software

Give AI agents structured access to real software