guilds.run

Most AI proxy platforms only provide you with a single result, making it impossible to see how the process actually unfolds. guilds.run does the opposite: it records every prompt, every tool call, and every change made to the system’s memory during each execution. In the event of any issues, it is possible to trace back to which specific proxy was used, which prompt led to the problem, and which step caused the error.

Compared to similar solutions based on code-based frameworks, it functions more like an interface-equipped team collaboration workspace. Teams can share tables, notes, and scripts, and data is retained across different runs while allowing for manual modifications. Within Studio, a “coach” role helps in reviewing past runs, drafting changes for review before they are finalized and released. Models can be combined using tools such as Claude, GPT, Gemini, Grok, as well as open-source models. Integration options cover platforms like Notion, Gmail, GitHub, X, etc., and permissions can be set down to the level of individual agents within a team. Additionally, there are ready-made team templates available for use on platforms like X and LinkedIn.

The project is open-source, and Docker can be used to run it locally with just one command. You need to provide your own model keys. The keys for the managed version and the public template library are currently marked as “coming soon,” and some connectors have not yet been released either. Those who expect to use the system out of the box will have to wait a bit longer.