We just finished a project where we swapped out AutoGen for Agno for a major client's internal support chatbot. I was brought in to help with the basic workflow setup.
The biggest surprise? How much simpler the state management was. In AutoGen, we had a lot of back-and-forth between the assistant and user proxy, even for straightforward Q&A against their knowledge base. With Agno, we basically defined a single agent with access to tools (search, ticket creation) and it just... worked. The conversation felt more linear and easier to log.
Also, the cost transparency was a huge win for the client. With AutoGen, the token usage across multiple agents chatting to each other got confusing. Agno’s single-agent approach made it way easier to predict and explain the monthly bill. We're now looking at using Agno for their customer-facing knowledge base too.
Curious if others have had similar experiences with complex workflows? Especially around integrating with existing ticketing systems like Jira.
newbie
Ask me in a year
I'm Brian, and I handle ERP and systems integration for a mid-size manufacturing company, roughly 400 employees. We run NetSuite as our core, and I was tasked with evaluating both AutoGen and Agno for a connected supplier portal chatbot that handles order status and RMA initiation.
Here's my breakdown from our pilot last quarter:
**Target Audience & Fit:** AutoGen felt architected for a lab or a very dedicated AI engineering team wanting to simulate complex, multi-actor scenarios. Agno was immediately more approachable for our internal devs who are integration-competent but not AI specialists. It's a better fit for IT or product teams embedding a single, capable agent into an existing service layer.
**Real Pricing & Cost Clarity:** AutoGen's open-source framework meant no direct cost, but the operational cost was opaque and ballooned quickly because you're running multiple LLM agents in a chain. Agno's pricing model, at least for us, was around $0.20 per 1k "workflow units" which roughly translated to 2-3 complex tool calls. For a 5k-user portal, our projection was $300-500/month, which was far easier to budget for than the variable, multi-agent token consumption we saw in testing.
**Integration & State Management:** This was the biggest difference. With AutoGen, we spent a week just tuning the conversation flow between the user proxy, assistant, and executor to avoid loops. With Agno, we defined one agent with access to our NetSuite SuiteTalk API and a document search tool. The state is inherent to the single conversation thread, which cut our development time for the POC by about 60%.
**Where Agno Clearly Wins (and Loses):** It wins on simplicity for deterministic, tool-using workflows like "answer from KB, then open ticket if needed." It's a straight line. It loses if you genuinely need competitive, debating agents for something like code review or design simulation. For our use case - a support chatbot that sometimes creates Jira tickets - the single-agent model was a feature, not a limitation.
Given what you've described - a straightforward internal support chatbot that primarily queries a knowledge base and sometimes creates tickets - I'd recommend Agno without hesitation. The cost transparency and linear workflow are exactly what you need. If your client later wanted to simulate a debate between a compliance officer and a support agent on every response, then you'd have a case for revisiting AutoGen.