Launching today
guilds.run
Design multi-agent systems that you can trust
61 followers
Design multi-agent systems that you can trust
61 followers
Open-source AI agent teams you can actually inspect. Debug every prompt, tool call, memory change, and agent interaction. Control permissions, edit context, and choose the model for each agent. Self-host for free. Stop guessing what your agents are doing.
Free
Launch Team / Built With
Framer AI AgentsDesign and publish professional sites with AI
Promoted
Hey Product Hunt! 👋
Getting AI to handle long-running tasks reliably takes more than good prompts. There’s a lot of system design involved: figuring out how agents should work together, anticipating edge cases, and being able to debug and tune the system when things go wrong.
That’s why we built guilds.run.
The goal is to make the whole system understandable, so you can trace how mistakes propagate between agents, fix the underlying causes, and see if your changes work.
We’re open-sourcing guilds.run because we’d like to see what other people built and share working team the solve real world use cases for long running agent collaboration
Coming next:
🧩 Recipe sharing: discover, reuse and share whole agent teams built by the community.
☁️ Hosted mode: the same product without managing your own infrastructure. Leave your email on guilds.run to get a seat when it opens.
If you like what you see, a ⭐ on GitHub helps a lot. We’re here all day, so tell us what breaks and what you’d build.
Hey PH! 👋
Super happy to be hunting guilds.run today.
Anyone who has run multi-agent setups knows the pain. One agent makes a wrong assumption, the next one treats it as fact, and 3 steps later you’re staring at a broken output with no idea where it went sideways. Most tools just hand you better prompts. guilds.run actually lets you see what happened.
What caught my attention:
🔍 Full visibility into every prompt, tool call, memory change and the group chat between agents
🛠 Studio reads your runs, finds the root cause and drafts a fix, and nothing goes live until you publish it
⚙️ Per-agent control over permissions, context and model
🆓 Open source and free to self-host, up and running with one docker command
Love that the team is building in the open.
Congrats @m4nuc and @vincent_cigale on the launch!
@vincent_cigale @kate_ramakaieva Thanks Kate. Great summary. We might keep it
As more non-developers start building with AI, we’ll need ways for them to understand and debug the systems they create. Interested to see where this goes.
@dipanshu_kushwaha5 thanks for your comment
You've got it , that's exactly why we built guilds.run :)
Hi there! I’m not technical at all, yet within a few hours, I set up automated prospecting on LinkedIn. It searches for and invites prospects, sends personalized messages, and follows up.
@vincent_cigale can we test it as a template?
Seeing where an assumption went wrong as it passes between agents sounds useful. Being able to follow that trail would help with figuring out what to fix. Congrats on the launch, Emmanuel and Vincent!
@oleksii_sekundant exactly. Thanks for the kind words Olek
3rd maker here! I have a PhD in cognitive science, and now I’m trying to understand what’s going on inside agent “brains.” 😅
Hope you like what we’re building!
@dadadel this guy, never miss an opportunity to show off his PHD 😝
@m4nuc My agent brain has a permission setting for PhD mentions. Default: always on 😅
Get started with:
@m4nuc So easy even a non-PhD can run it 😜
@m4nuc @dadadel 😅