I wrote a post on creating "scalable research tooling for agent systems" and I'm also releasing the companion MCP server which lets you do autonomous Frida instrumentation on Android. Details in thread 👇📲🪝

Over the last few years I have done quite a bit of AI tooling and I thought it was a good time to share some of my learnings. I'm really interested in feedback from other people who are working on similar things! Blog: GitHub:

Frida is such an amazing tool, people shouldnt sleep on mobile as a viable initial access method

Frida + MCP is a smart combo. Letting an agent drive dynamic instrumentation on Android opens up some wild analysis workflows. Bookmarking the blog post too.

this is exactly the kind of tooling the security research space needs. frida instrumentation through mcp opens up so much automation potential for android analysis. curious how you handle state management when the agent is running multi step hooking workflows?

I think this is a question that should be answered by the orchestration layer, how planning happens, how to track what has been done/is being worked on, if a piece of work is a keystone that should be persisted for other agents to use, how it fits into the larger context of the current objective. Maybe later I will write a post about this as well. Models make this less hard with each update because of natural market pressures to make them better at planning and handing out tasks to sub-agents.

would definitely read that post. the keystone persistence idea is interesting, most orchestrators i've seen treat each agent run as stateless which loses so much accumulated context. a shared artifact layer between agents feels like the missing piece.