⚠️ Affiliate Disclosure: This post is not an affiliate associated review. It may become one in the future and this disclosure will reflect that. This does not affect my reviews and I only recommend tools I genuinely believe in.
guilds.run is an open-source platform for running a team of AI agents, built by the French company Cigale (the GitHub organisation is CigaleAI) and launched on Product Hunt on 13 October 2026 — today, as I write this. I haven’t tested guilds.run — this is an information guide for developers weighing it up, not a review. Everything here comes from guilds.run and the Product Hunt listing, fetched on 13 October 2026.
What Is guilds.run and What Does It Do?
The tagline is “Create multi-agent workflows that you can trust”, and the whole product is built around one complaint: multi-agent systems are black boxes. guilds.run’s answer is inspectability. Per the site, each run keeps the exact prompt every model saw, every tool call with its result, every memory change, and a record of which agent wrote each note. The makers describe it on Product Hunt as “open-source AI agent teams you can actually inspect” — you can debug prompts, tool calls and memory, control permissions, edit context, and choose a different model for each agent.
It’s self-hosted: the quick start clones the GitHub repo, runs docker compose up --detach --wait, and serves the app in your browser at 127.0.0.1:8080. A hosted version is listed as coming soon.
How Does guilds.run Work?
You organise agents into teams (“guilds”). A Studio assistant reviews your live team and drafts fixes, but “Nothing changes until you publish”, per the site. Agent notes, memory and scripts are versioned and editable, with each change attributed to the agent or step that made it. Workflows can be bundled into shareable “recipes” — agents, tasks, tables and schedules in one file — with a public directory marked coming soon. Permissions attach connectors at the organisation, guild or single-agent level: “The next run only offers what you turned on.” Each agent can get its own computer — a terminal, a workspace and optionally a browser that stays logged in.
On models, guilds.run is bring-your-own-keys: Claude, GPT, Gemini, Grok and open-weight models such as Qwen and Llama, accessed through OpenRouter, DeepInfra, Ollama or your own provider keys. Platform-supplied keys on the hosted version are listed as coming soon.
How Much Does guilds.run Cost?
| Self-hosted | Free — the Product Hunt listing shows “Free” and the site says “Open source, on your machine”; you pay only your own model API usage |
| Hosted version | Coming soon; Lite, Standard, High or custom instance tiers listed with CPU/RAM figures but no prices published |
| Model usage | Your own keys (OpenRouter, DeepInfra, Ollama, direct providers) — billed by those providers, not by guilds.run |
On Product Hunt, one of the makers said there are “no official prices yet” for the hosted version but that they want “a very cheap entry price”. So the honest summary: free to run yourself today, hosted pricing unknown.
What Are Developers Saying About guilds.run?
Too new for real-world feedback — it launched the day I’m writing this. The only signal so far is the launch itself: when I checked on 13 October 2026, the Product Hunt listing showed 184 upvotes and a day rank of #4, which is decent launch-day traction but tells you nothing about how it holds up in use. I found no Hacker News or Reddit discussion yet.
Who Is guilds.run For?
On paper, guilds.run is for a developer or small team that wants multi-agent automation but refuses to trust what they can’t audit — the prompt-level run history and scoped permissions are the draw. You need to be comfortable self-hosting with Docker and managing your own model keys and costs; there’s no sign-up-and-go option until the hosted version ships. If you’d rather build agent teams in code, CrewAI, LangGraph and AutoGen are the established names in this space — mentioned as alternatives only, not a comparison I can source. One homework item before committing: the site doesn’t state the open-source licence, so check the CigaleAI/guilds repo on GitHub first.
guilds.run: Pros & Cons On Paper
Pros
- guilds.run documents full run inspection: exact prompts, tool calls with results, and attributed memory changes
- Open source and free to self-host, with a two-command Docker quick start
- Model-agnostic by design — Claude, GPT, Gemini, Grok and open weights via your own keys, per agent
- Permissions scoped to organisation, guild or single agent (documented, not tested)
- Studio changes stage until you publish them — nothing edits your live team silently
Cons
- guilds.run is self-host only today — the hosted version and its pricing don’t exist yet
- The open-source licence isn’t stated on the website
- Several pieces are marked coming soon: the recipe directory, some connectors, platform keys
- You carry all model API costs and key management yourself
- guilds.run launched today — zero production track record
Is guilds.run Worth a Look?
This isn’t a scored verdict — I haven’t run guilds.run. What I’ll say on paper: the inspectability pitch addresses the most common and legitimate complaint about multi-agent frameworks, and because it’s free and open source, the cost of checking whether the reality matches the pitch is one Docker command and your own API keys. The sensible caution is the usual one for a day-old project: coming-soon labels outnumber shipped hosted options, and the licence needs checking before any commercial use. If you’re already sceptical of black-box agent teams, this is the launch in this batch I’d watch most closely.
Used guilds.run? Tell Me How You Find It
I don’t test developer tools — you’re the expert here. If you’re using guilds.run, leave a comment: what works, what doesn’t, and what you’d choose instead. Honest developer experience is worth more than anything I can write from the vendor’s pages, and it helps the next dev weighing it up.
Leave a Reply