AI news · October 13, 2026
guilds.run opens inspectable AI agent teams as open source
guilds.run launched on Product Hunt on October 12 with a Docker setup and AGPL code; builders can inspect every prompt, tool call and memory change in a local agent team.
- setup commands in the README
- 3
- software licence
- AGPL-3.0
- local operator UI
- 1
guilds.run is an open-source framework for standing teams of AI agents. Each agent has a role, model and tool set; the team shares a chat, notes and tables. The README says every run keeps the prompt the model saw, each tool call and every memory change. That makes it possible to inspect how one bad assumption moves through a team instead of looking only at the final answer. The project is AGPL-3.0 and can be started with Docker after cloning the repository.
This is a practical response to a common multi-agent failure: one worker makes a wrong assumption and later workers treat it as fact. The cost is that the operator must manage Docker, provider keys, permissions and updates. The hosted edition is separate, so the self-hosted route is the immediate one. Run a small research or file-triage guild, inspect the trace, and compare its useful output with one well-prompted agent before adding more roles.
What you can do with it
Clone the repository, run setup.sh and docker compose on a test machine, then connect one model provider. Give the first guild a bounded task with a known answer. Inspect the prompts, tool results and memory changes, and remove any agent that adds no useful work before trying a longer routine.
Our take
guilds.run has a clear product idea: visibility and control are part of the agent team, not an afterthought. The open code makes a local test possible, but the project is young and the hosted option is not yet open. Its early value is as an inspection tool for agent design, not as a reason to replace every simple workflow with a team.
Source: Cigale AI / Product Hunt — Made With Models writes the brief; the reporting is theirs.