Performance meets scale: a month of relentless innovation
Last month was a big one here at Agno. From groundbreaking async database support to enterprise-grade guardrails, our community shipped features that don't just move the needle; they redefine what's possible with agentic AI. With 13+ major releases, 100+ contributors, and a surge of production-ready tooling, this month was all about performance, scale, and developer empowerment.
Plus, we welcomed 130+ new community members to our Discord, a clear sign that Agno’s agentic AI movement is accelerating.
Whether you're building conversational workflows, managing multi-agent orchestration, or deploying agents at enterprise scale, this roundup has something to make your work faster, more powerful, and, hopefully, more fun.
New releases & features
v2.1 - v2.2.11: The performance era
Over the past two months, we’ve maintained a relentless pace of improvements. Here’s what we shipped:
Response caching (v2.2.2)
Speed up development and cut costs with LLM response caching. Simply set cache_response=True on your model class and watch redundant API calls disappear. This is perfect for improving efficiency in testing and iterative development scenarios.
Async database support (v2.2.3)
Agno now fully supports asynchronous database operations across MongoDB, SQLite, and PostgreSQL. This is a critical infrastructure upgrade that eliminates blocking I/O, which often kills performance in production. The result is truly concurrent agent execution at scale and significantly better performance for your workloads.
Conversational workflows (v2.2.6)
Workflows aren't just for batch processing anymore. With conversational workflow support, you can now build chat-like experiences directly into your workflow steps. This bridges the gap between rigid automation and flexible agent interactions.
Claude Skills & Context management (v2.2.11)
Native support for Claude's advanced capabilities, including its reasoning, code execution, and tool interaction skills. A crucial benefit in this release is the addition of Claude's context editing features. This allows Agno to automatically control growing context windows by removing old tool and reasoning blocks, which is essential for managing long-running agent sessions efficiently.
Knowledge search filters (v2.2.12)
Advanced metadata-based filtering with a new DSL (EQ, IN, GT/LT, NOT, AND, OR) makes knowledge retrieval surgical. This feature is currently supported in PGVector, with broader vector database support coming soon, and includes improvements to agentic filtering for smarter retrieval.
Batch embeddings (v2.1.0)
Stop making one API call per embedding! We now support batch embedding, which significantly speeds up knowledge base ingestion and dramatically reduces API overhead and costs. Updated documentation is now available.
Guardrails & security (v2.1.0)
Production AI demands safety layers. Agno now includes comprehensive, built-in guardrails to help you build responsible AI without reinventing the wheel. Features include:
- Prompt injection detection
- PII detection (SSN, credit cards, emails, phone numbers)
- OpenAI moderation checks
- Custom guardrail support
Pre/post hooks (v2.1.0)
Configure custom hooks that run at the start or end of agent/team runs. Perfect for input/output validation, transformation, logging, or custom telemetry.
Parallel tools toolkit (v2.2.11)
External data retrieval is now blazingly fast for AI agents. We have optimized web search and content extraction, supporting both direct API integration and MCP server support.
Notion toolkit (v2.2.6)
Give your agents read/write access to Notion pages. Use cases range from documentation automation to knowledge management workflows.
RedisVL vector database (v2.2.7)
Redis joins our vector database lineup! You can now use Redis for knowledge storage with full asynchronous support. This integration is perfect for teams already invested in the Redis ecosystem.
vLLM embeddings (v2.2.7)
We now support local and remote vLLM models for generating embeddings. This allows you to run embeddings entirely on your own infrastructure, providing independence and eliminating external API dependencies.
A2A interface (v2.1.2)
Expose your agents, teams, and workflows in A2A-compatible format via the AgentOS. Enables interoperability across agent frameworks and platforms. Here’s the full documentation.
Input validation & model strings (v2.2.6)
Agents and teams now support input schema validation on AgentOS. Plus, you can define models as simple strings: model="openai:gpt-4o". Small changes, big developer experience wins.
Community projects & showcases
Our community continues to build incredible things. Here are the standout projects from the past two months:
1. Project spotlight: agno-go by Alexandre E. Souza
The Golang port of Agno just keeps getting better. Recent releases added:
- MCP support — Full Model Context Protocol integration (demo video)
- MaxToolCallsFromHistory — Manage context size and reduce tokens
- Input/Output Schema support — Bringing Python Agno's structured data capabilities to Go
This isn't just a port. It's a full-featured framework for Go developers who want agentic AI without switching languages. Check out the repo.
2. Customer support agent by Bobur Umurzokov
A single-widget solution for AI-powered customer support that actually remembers. Built with Agno + Memori OSS, this agent:
- Uses Agno for knowledge retrieval and agentic workflows
- Leverages Memori to persist facts and user context
- Installs on any website with one line of code
- Brings your own data from URLs, documents, and websites
Perfect for teams wanting personalized support without building from scratch. Try it out.
3. FormPilot by Shivam Goyal
An AI-powered web form automation framework that eliminates repetitive data entry. FormPilot can:
- Detect and map input fields on any webpage
- Validate and format user data intelligently
- Auto-fill and submit forms
- Integrate with APIs or local datasets for dynamic input
- Smart Link Extraction using MCP tools
Built with modularity in mind, developers can extend FormPilot for custom workflow automation. GitHub | Portfolio
4. NewsVerse by Darsh Vaishnani
A personalized news platform that learns what you care about, not just what you click on. NewsVerse uses Agno Framework for:
- AI-powered article analysis (professional + kid-friendly summaries)
- Fact-checking via web search
- Sentiment analysis and entity extraction
- Content matching using embeddings and cosine similarity
- Personalized delivery via web app or WhatsApp (using Twilio)
The stack leverages Agno's multi-agent architecture with built-in API failover (Gemini → Groq), enabling smooth development despite rate limits. Presentation | GitHub
5. Briefly by Rao
Record voice notes, thoughts, or meeting moments from your wrist or phone and remember them forever. Briefly:
- Connects to your calendar and sends pre-meeting notifications
- Records meetings and extracts context
- Labels Gmail emails and drafts replies using your meeting context
- Uses Agno for the entire agent backend
6. Building your first AI agent tutorial by Sudarshan Koirala
A beginner-friendly video tutorial walking through creating your first AI agent with Agno. Perfect for developers new to agentic AI who want a practical, hands-on introduction. The tutorial covers agent fundamentals, setup, and basic workflows in an accessible format. Watch on YouTube, the ideal starting point for Agno newcomers.
7. AgentOS Revolution tutorial by TechieTalksAI
A comprehensive video tutorial showing how to build autonomous AI teams with Agno's Agent as a Service architecture. The tutorial covers AgentOS setup, team coordination, and deployment patterns. Watch on YouTube: great for developers new to multi-agent systems.
8. Universal support bot by Raghavender Reddy Grudhanti
An AI-powered enterprise support system that eliminates the frustration of traditional support portals. Built with Agno, OpenAI, and Streamlit, this bot lets users report issues in plain English—no forms, no categories, no portal navigation. Just conversation.
The bot automatically understands user intent and routes requests to the right system:
- Salesforce for billing and account issues
- Jira for product or application problems
- ServiceNow for IT and access requests
A perfect example of using Agno to unify multiple enterprise systems behind a single conversational interface. Read the full article.
9. vLACQ stack by Kameshwara Pavan Kumar Mantha
A production-ready, composable RAG architecture was unveiled at the vLLM Meetup Hyderabad. The vLACQ Stack combines:
- vLLM — High-performance LLM inference
- LlamaIndex — Data framework and indexing
- Agno — Agentic orchestration and workflows
- Chonkie — Semantic chunking
- Qdrant — Vector database
This isn't just another RAG framework - it's an enterprise-grade system designed for both agentic applications and standalone RAG pipelines. From handling complex financial tables to multimodal reasoning, vLACQ delivers reliability, scalability, and precision at every step.
The stack powers vision-language intelligence, semantic retrieval, and agentic reasoning in production environments. Read the architecture deep-dive.
Contributor shoutouts
Last month brought 100+ contributors across 13+ major releases, showcasing the power of open-source collaboration. Special recognition to our community contributors:
Top code contributors
- @kredde (GitHub) — Custom events in AG-UI interface integration
- @harrytran001 (GitHub) — DynamoDB schema fixes and serialization improvements
- @SaiPoornaChandraPrakash (GitHub) — Google Drive toolkit implementation
- @jtalmi (GitHub) — Tool post-hooks for failure handling and Claude context management
- @mickvav (GitHub) — FileTools relative path handling and file operations enhancements
- @vincentkoc (GitHub) — Opik telemetry documentation
- @fatelei (GitHub) — Security vulnerability fixes for h11 CVE
- @renweizhukov (GitHub) — Google Calendar unit tests
Feature implementers
- @devalexandre (GitHub) — MCP binary support and agno-golang continuous improvements
- @s-smits (GitHub) — Default Claude model updates to Sonnet 4.5
- @kwojcik-blockether (GitHub) — Session renaming via SQLiteDb fixes
- @antipopp (GitHub) — AG-UI state integration improvements
- @gkorland (GitHub) — User ID handling in memory deletion methods
- @Shreya860 (GitHub) — Workflow updates for package management
- @KarthikeyaKollu (GitHub) — PDF reader refactoring for efficiency and code reuse
- @zhangwfjh (GitHub) — Code cleanup and optimization
- @10Ala10 (GitHub) — PPTX reader support
- @rbs333 (GitHub) — Redis VectorDB support
- @peter-mw (GitHub) — Editable summary message prompts
First-time contributor
In the last few weeks, we've extended a warm welcome to 40+ first-time contributors, including @simonkagwi (Google Sheets service accounts), @souvikinator (WhatsApp session improvements), @MatthiasJostLpz (documentation fixes), @ahmedsamirio (model response handling), @niradler (async database improvements), @lacia-hIE (media handling fixes), @alec-drw (knowledge base async updates), @ron-42 (async session summary methods), @raphaelrodrigues (parent run ID fixes), @bhatt-neel-dev (workflow session improvements), and many more. Your PRs, bug reports, and feature requests are the lifeblood of this project.
Community MVPs
- @aberk — Tireless debugging help, detailed bug reports on async databases, tool metrics, and Claude integration edge cases
- @Sebas — Deep technical questions that pushed the boundaries of Claude integration and beta header handling
- @DennisK — Workflow testing, type safety improvements, and thoughtful UI feature suggestions
- @yathaarth — In-depth investigation of run_context behavior and message handling patterns
- @Griflet — Advocating for critical features like HITL in teams and agent pause/resume capabilities
- @Omar — Session summary, edge case debugging, and real-world production feedback
What's next
On the roadmap
- HITL (Human-in-the-Loop) for teams — One of the most requested features.
- RBAC (role-based access control) for AgentOS — A core security and provisioning improvement currently underway.
- Enhanced MCP tooling — OAuth support, better error handling, and lifecycle management improvements.
Get involved
We're always looking for contributors! Areas where you can help:
- Documentation — Help new users get started faster
- Integrations — Build connectors for your favorite tools and services
- Cookbooks — Share real-world patterns and use cases
- Bug reports — Production feedback makes Agno better for everyone
Check out open issues or join the Discord to connect with the team.
Join the movement
This month continued to prove that Agno is defining the agentic AI revolution.
Whether you're prototyping your first agent or deploying multi-agent systems at scale, Agno gives you the tools to move fast without breaking things. And with a community this vibrant and growing this quickly, you're never building alone.
- ⭐ Star Agno on GitHub
- 💬 Join our Discord
- 📖 Read the docs
- 🚀 Share what you're building. We want to feature YOUR project next month.
See you in the next Agno Community Roundup. Keep building, keep shipping, keep pushing the boundaries of what agents can do.
🧡 Team Agno