What is this?
Agno is a framework and runtime for building, running, and managing agent platforms: build agents with the SDK, serve them via the AgentOS runtime, and manage everything from a web UI.
Agno is a framework and runtime for building, running, and managing agent platforms: build agents with the SDK, serve them via the AgentOS runtime, and manage everything from a web UI.Is Agno a framework, a server, or a hosted app?Both halves: a Python SDK for building agents, plus AgentOS — a self-hosted runtime that serves them as a service with a REST API, a control plane, and a web UI.
Both halves: a Python SDK for building agents, plus AgentOS — a self-hosted runtime that serves them as a service with a REST API, a control plane, and a web UI.How do I get a first platform running?Hand the quickstart prompt to a coding agent like Claude Code, Cursor, or Codex: it spins up the stack locally via Docker — REST API, Postgres, MCP server, control plane. Or hand-code your first agent in 20 lines.
Hand the quickstart prompt to a coding agent like Claude Code, Cursor, or Codex: it spins up the stack locally via Docker — REST API, Postgres, MCP server, control plane. Or hand-code your first agent in 20 lines.Do I need Docker and a coding agent for that?Docker for the local run and Postgres for storage; the coding agent is optional, since the README also shows building your first agent by hand in 20 lines of code.
Docker for the local run and Postgres for storage; the coding agent is optional, since the README also shows building your first agent by hand in 20 lines of code.Where can I deploy it once it works?Anywhere that runs containers: locally via Docker, with starter templates for Railway, AWS, GCP, Azure, Fly, Render, Modal, and Helm — only the deploy scripts differ.
Anywhere that runs containers: locally via Docker, with starter templates for Railway, AWS, GCP, Azure, Fly, Render, Modal, and Helm — only the deploy scripts differ.Does it send any of my data home?One telemetry event per agent run, so the team knows which model providers to prioritize; prompts, messages, and outputs are never sent, and you can turn it off with the AGNOTELEMETRY=false setting.
One telemetry event per agent run, so the team knows which model providers to prioritize; prompts, messages, and outputs are never sent, and you can turn it off with the AGNOTELEMETRY=false setting.What does AgentOS give me out of the box?A production REST API with 50+ endpoints (SSE and websockets), storage for sessions, memory, knowledge, and traces in your own database, an MCP server, JWT-based RBAC with multi-tenant isolation, OpenTelemetry tracing with run history and audit logs, and cron-based scheduling and background jobs with no extra infrastructure. There are also 100+ pre-built toolkit integrations like GitHub, Slack, and Postgres.
A production REST API with 50+ endpoints (SSE and websockets), storage for sessions, memory, knowledge, and traces in your own database, an MCP server, JWT-based RBAC with multi-tenant isolation, OpenTelemetry tracing with run history and audit logs, and cron-based scheduling and background jobs with no extra infrastructure. There are also 100+ pre-built toolkit integrations like GitHub, Slack, and Postgres.Where can users actually talk to my agents?Out of the box your agents can be exposed via Slack, Telegram, WhatsApp, Discord, and the AG-UI and A2A interfaces, so you don't have to build those chat integrations yourself.
Out of the box your agents can be exposed via Slack, Telegram, WhatsApp, Discord, and the AG-UI and A2A interfaces, so you don't have to build those chat integrations yourself.How does it stop an agent from doing something risky?Runs can pause for human approval before sensitive steps, and individual tools can be blocked so they require admin confirmation before executing.
Runs can pause for human approval before sensitive steps, and individual tools can be blocked so they require admin confirmation before executing.Why own the whole agent stack instead of using a hosted service?The README's pitch: keep control of your data, memory, and security posture with JWT-based RBAC, store everything in your own database, and turn the platform into a learning loop with simulations and usage data.
The README's pitch: keep control of your data, memory, and security posture with JWT-based RBAC, store everything in your own database, and turn the platform into a learning loop with simulations and usage data.What are similar projects?turnstone, evenfire, hive, voltagent, agent-framework, DocsGPT, oh-my-subagents, trigger.dev
turnstone, evenfire, hive, voltagent, agent-framework, DocsGPT, oh-my-subagents, trigger.devWhich topics is it in?background-jobs, human-in-the-loop, rbac, telegram, agent-framework, observability, postgresql, sdk, multi-agent, agents, rest-api, cloud
background-jobs, human-in-the-loop, rbac, telegram, agent-framework, observability, postgresql, sdk, multi-agent, agents, rest-api, cloud