Twitxy Saved

Agno

This account is verified because it's an official organization on X. Learn more

@AgnoHQ

Agno is the programming language for agentic software.

Build, run, and manage multi-agent systems at scale.
agno.com git.new/agno Joined July 2022
182 Following 13K Followers 1.9K Posts

Latest posts 5

Agno @AgnoHQ
Only 11% of companies have scaled AI agents across their workflows. Most of the other 89% are paying for AI and still doing the work themselves.

So where's the gap? Look closely at companies that call themselves AI-native, and you'll find disconnected agents handling one task at a time. An agent summarizes a ticket or drafts a reply, then hands it back. A person still decides what happens next and keeps the work moving manually.

In truly AI-native companies, AI does more than help people finish individual tasks faster. Agents own areas of the business and run workflows from start to finish, including deciding what work needs to happen and when.

In our new post, we break down why so many companies stall before becoming AI-native and how to start closing the gap with one workflow you already do.

We'll link the blog in the first reply.

Where does your company fall within the four levels of AI maturity?
Agno @AgnoHQ
Your customer asks an AI assistant to do the work instead of logging into your product. A redesign won’t fix that. You can’t drag people back to a UI they’ve stopped opening.

So if an assistant is going to do the work, make it yours: your product’s logic, your permissions, your brand on the answer, and your usage data.

Slack is a good place to start because your customers are already there all day.

We published a guide on connecting an Agno agent to Slack. All it takes:
• A Slack app manifest
• Two environment variables
• A few lines of code to add the interface to AgentOS

We'll put the link to the guide in the first reply.
Agno @AgnoHQ
Not every AI project needs an agent. For a lot of jobs, a workflow is the better choice.

The difference comes down to who decides what happens next.

In an agentic workflow, that’s your code. You define the steps and the branching rules, and a model handles specific tasks inside that structure.

An AI agent hands control to the model. You give it a goal and some tools, and it works out its next move based on what it finds.

So, when should you use each?

•Workflows fit repeatable processes where no step can be skipped, like approvals or invoice processing.
•Agents are better for open-ended jobs where the path isn’t known up front, such as troubleshooting or answering customer questions.
•Often you’ll want both. Let the workflow enforce the rules and hand the judgment calls to an agent.

Agno treats both as first-class building blocks, so you can combine them in one system instead of forcing a single approach onto everything.

We put together a guide on how the two differ and when to pick each, with Agno code examples for building both. You’ll find it in the first reply.

Where are you running an agent today that would work better as a workflow?
Agno @AgnoHQ
Your AI feature needs a price, but you probably don’t even know what it costs to run yet. Even a good average can mislead you.

Say an in-product copilot gets 30 requests a day from each seat. On GPT-6 Sol, that’s about $10.56 per seat per month, which a $20 AI add-on can carry.

Then a few of your seats start using it a lot more. At 150 requests a day, a seat costs you $53, more than twice what it pays you. A flat add-on charges that seat the same $20 as everyone else, so you pay the other $33. Get enough seats like that and the add-on stops making money.

So before you pick a number, find out what your heavy users cost. We wrote a guide that walks through the math:
•Cost breakdowns you can rerun with your own numbers
•Four pricing models and when each one breaks
•Allowances and overages for your heaviest users
•Tracking cost per customer from your run data

Full guide in the first reply.
4 replies 2 reposts 10 likes 773 views View on X
Agno @AgnoHQ
Agents shouldn’t need a prompt to know what to do next, and coding agents shouldn’t lose their work when a host goes down. In October, we shipped 13 releases and made huge strides on both fronts.

Our community was busy too, with 35 contributors merging 41 PRs throughout the month.

There’s a lot to cover, so here are the highlights:

CompanyOS gives agents ownership of the work that never makes the roadmap. Define your Products and the Sources their agents should consult, and routines propose Projects and Tasks. Once your team approves the work, agents carry it out and bring finished Tasks back for review.

CodexAgent joins ClaudeAgent in Agno, so Codex and Claude Code both run with sessions and the AgentOS API. Over the month, they also picked up background runs, typed output, approvals, and support for teams and workflows.

Sandboxes give each Claude Code or Codex session its own computer. Agno checkpoints the workspace and conversation files after every turn, so if the container dies, you lose nothing past the last completed turn. Any replica can then pick up the session.

Compaction keeps long sessions inside the context window. When a provider rejects a request as too long, Agno summarizes the older turns with the agent’s own model and retries.

Agents also got a durable filesystem in their own database, so they can write down a decision and find it again in a later session. And AgentOS can now manage users and roles in your database too.

The community was busy too. A few projects we loved:

•Millie OS by @chaseschwalbach at Millie: 20+ Agno agents draft patient replies and review items at a healthcare clinic, while a rating agent flags weak conversations and drafts prompt fixes for the tech team.

•PitwallIQ by Muhammad Ikhwananda Rizaldi: an F1 strategy dashboard that pairs FastF1 telemetry with an Agno multi-agent engine to evaluate undercut windows and tyre cliff warnings.

•Vinay Chaudhari built a Laya router for Agno teams that replaces the LLM call behind each routing decision with a single forward pass through Laya’s decision model. Vinay’s numbers put it at under 35 ms with no LLM tokens.

October MVPs: Ilya Bogin for KeenableTools (@KeenableAI), @iuditg for the 60db voice tools, @neutr0n420 at Anakin (YC S21) for AnakinTools, @devYRPauli for three fixes across YouTube and Mongo tools, and John Cole (@siliconimist) for the ApifyTools upgrade to apify-client 3.x. Thank you for helping make Agno better.

And thanks to everyone who built with us this month. Full roundup and contributor list in the comments.