Launched this week
Ship AI agents to production in hours. Opengeni gives you sessions that recover from failures, isolated sandboxes, 100+ integrations, human approvals, and visibility into every step and dollar spent. Focus on your agents. We handle the infrastructure. Self-host for free, or get started on our cloud. For this launch, the first 100 users get $100 in cloud credit. Promo-code: PRODUCTHUNT100
Hi Product Hunt.
For the past two years, we’ve been running AI agents against real production cloud infrastructure at CloudGeni. We learned the hard way that getting an agent to work once is only the beginning. Keeping it running, giving it somewhere safe to execute, connecting users’ accounts, getting human approval, and understanding what happened—and what it cost—is where the work adds up.
We kept rebuilding that infrastructure. So we pulled it out and open-sourced it.
That’s Opengeni: agents in your product, infrastructure out of the box.
It gives developers the pieces they need to ship agents, already wired together:
🔄 Sessions that survive errors, with a replayable history so you can reconnect and pick up where you left off.
🖥️ Sandboxes for agents to do the work, with credentials handled and memory built in.
🔌 Your APIs and your users’ apps, connected. Bring an OpenAPI or GraphQL spec, or an MCP server, plus 100+ integrations with built-in OAuth.
✋ Human approvals before risky actions, so people stay in control.
🔍 Visibility into every step, token, and dollar, with built-in dashboards and run traces.
🧩 A workspace for each customer and the freedom to choose your models.
Getting it into your product takes three steps: add your API key on the server, mount a handler using your existing authentication, and show the agent in your app—with your own UI or our React components.
You can also try the same sessions in the Opengeni app before embedding them in yours.
Run it on our cloud or yours. Opengeni is Apache-2.0 and free to self-host. Our managed cloud is model cost + 5%, with no platform or seat fees. For this launch, the first 100 users get $100 in cloud credit.
We’re in early public preview, and we’d love your feedback: what’s the first task you’d want an agent to handle inside your product—and what would you need to trust it with that task?
Try the cloud · Read the docs · Self-host on GitHub
We’ll be in the comments all day. Thanks for checking us out!
@alchemist_studio what kind of performance impact do you notice on Postgres when scaling up to thousands of active sessions?
@alchemist_studio @priya_kushwaha1 Hey Priya! Here is how we handle it: idle sessions are basically free. No workflow stays open and nothing polls Postgres for them, they're just rows sitting there.
The real work happens when a turn is actively streaming. We batch token writes (every 33 ms or 50 events), and those batches don't touch the main session row, so busy streams don't all fight over it. Live viewers get a NATS nudge and read from Postgres, and the API is stateless, so you can just add more replicas.
If you're planning something big, open an issue and we'll dig in with you: https://github.com/Cloudgeni-ai/opengeni/issues
@alchemist_studio congratulations on launch buddy
@savan_kharod1 Thanks a bunch! @jorgensandhaug was the originator of the product almost 4 months ago. Kudos also to @benyhh another lead dev on the product and to the rest of the team as well. 🚀
@alchemist_studio super nice that you offer the open source / self-managed version! :) congrats on the launch
@alchemist_studio congrats on the launch! and best of luck today
@alchemist_studio Congrats on the launch my man! Curious who the ideal users are for the product: SaaS teams embedding agents in their product / internal platform/DevOps teams?
Would love to hear who's getting the most out of it so far but also where the team is hoping to hit scale.
Congratulations on the launch team!
Was curious if Open Geni is suitable for serverless deployment, because as I can see it currently supports SSE, and for long running multi agents, I would prefer an efficient streamable connection to avoid extra cloud costs. Is that already worked upon or any open tickets for the same?
Regardless, I think Open Geni being model agnostic and open sourced helps a lot!
@prastik Thanks Prastik, appreciate it!
Good news: long multi-agent runs don't need an open connection at all. You send a message, get a 202 back, and the agent keeps working on a background worker. Close the stream whenever you want, nothing gets cancelled. Sub-agents report back through Postgres, not through your connection.
If you don't want to stream, use our webhooks (turn.completed, session.requiresAction, etc.) and fetch the events when you need them. If you do stream, it picks up where you left off. And if you meant MCP, our MCP endpoints already use Streamable HTTP.
On serverless: the core itself doesn't scale to zero today. The workers are long-running and we deploy on Kubernetes or Native Cloud. Your own backend can totally be serverless though, since the SDK runs on edge runtimes. One tip: don't proxy the SSE stream through a function, it'll stay open and cost you. Webhooks are the cheaper way to go.
More here: https://docs.opengeni.ai/integrate/background-agents
No public issue on this yet, feel free to open one! https://github.com/Cloudgeni-ai/opengeni/issues
Congrats! Looks interesting! How do you see Opengeni being different from teams building their agent infrastructure directly with frameworks like LangGraph or the cloud providers?
@hakon_hogetveit1 Hey Håkon, and thanks! Teams can absolutely build this themselves, but from our own experience, building AI infrastructure is hard, and there are lots of considerations that only become apparent once you get started.
With Opengeni, teams can use infrastructure that’s already in place and focus on their actual products. Since it’s open source, they can also self-host it and adapt it however they like, saving a lot of time along the way!
@hakon_hogetveit1 4600 commits in 4ish months. 300 PRs just last night 😅. And this is just the start. I think Opengeni is at 10% of its potential, and what it can become in the future, even tho it has feature parity or beyond its competitors. Last time I checked. 1.5 Million lines of code was added just the last 30 days. And it was already 1m loc. I would not build or manage this my self. "Benny" and "Jorge" are going Rainman mode on this, and its perplexingly hard stuff to get right. Use this and focus on your end product instead. Its opensource so you can shape it to your need over time as well. Also the team ships feature requests really fast. You can also switchout the harness for other harnesses pretty easily, if needs change. Upwards and onwards. 🚀 Love vouch!
@alchemist_studio awesome stuff. 🔥
@hakon_hogetveit1 @alchemist_studio 300 PRs in a night reads as a red flag to me, not a strength. No team can meaningfully review that volume...
I'm also unsure about the managed cloud. Most teams will run closed models from OpenAI or Anthropic, and both now have their own managed agent offerings in public beta (Agents API and Claude Managed Agents), with no platform fee on top of tokens. I see the case for self-hosting, EU data residency (though OpenAI has EU residency and Anthropic the same via GCP) and being model-agnostic
@simona_o_neill3 Thank you! We've seen teams that already have existing saas product, struggling with how to think about architecting and introducing capable agents, without breaking existing boundaries between users, organizations and what is possible to do in the product. Our solution helped them to start very simple and at the same time have full capability for scaling it beyond POC
Congrats on the launch!!!
Gotta deploy OpenGeni so i can use OpenGeni to deploy OpenGeni.
I'm exited to see what you guys do in the space. This looks a lot more mature than the alternatives I've evaluated so far. It's the kind of thing that separates your product from "chatgpt api wrapper" to "yes, we run agents and we can tell you exactly how".
@internetyngve Haha, we were literally talking about this last week! In theory you can run Opengeni inside Opengeni inside Opengeni... all the way down, until the RAM gives up. 😅
Thanks a lot, that last line really made our day. Being able to show exactly what an agent did and can we trust them with important stuff is big reason we built this.
@internetyngve We use opengeni to build opengeni :D Btw suuper easy to deploy to any cloud or machine, e.g. your own vm behind tailscale or whatever, recommend trying that!
@jorgensandhaug fiddling with the helm values as we speak!
@internetyngve lmk if you hit any problems, and ill help you asap
Seeing where every penny goes is a wonderful feature. Is there an option to put a limit on how much budget each agent and session uses, and for the run to stop once the limit is reached?
Btw, Congratulations @davletdz , @iuliia_petryshyn_thuen & team Opengeni 🎊✌️
@iuliia_petryshyn_thuen @aymi_malik Sounds like a great suggestion for https://github.com/Cloudgeni-ai/opengeni
We already track everything including cost per user and session and we have limits and guardrails in place, so combining two will be fairly simple :)
Love that you are launching this to a wider audience!
@We Are Learning have used @Opengeni for the last year, and I keep referring to @Opengeni for startups and scaleups that needs to be secure (SOC2 or similar) while using minimum bandwidth thinking about it.
@atlemk Appreciate support Atle! It's been amazing to work alongside great companies like @We Are Learning and make meaningful impact for growing scaleup needs.