diff --git a/.gitignore b/.gitignore index 1c19dd76..42b88f35 100644 --- a/.gitignore +++ b/.gitignore @@ -8,12 +8,13 @@ linkchecker-out.txt .DS_Store .mintlify-last .cursor +.claude +.agent +.envrc agno projects specs -.claude -.envrc agno-docs-style/styles/alex agno-docs-style/styles/write-good agno-docs-style/styles/Google diff --git a/agent-os/introduction.mdx b/agent-os/introduction.mdx index ac4e9ec5..3b2314f7 100644 --- a/agent-os/introduction.mdx +++ b/agent-os/introduction.mdx @@ -4,17 +4,15 @@ sidebarTitle: "Introduction" description: "The runtime and control plane for multi-agent systems." --- -**AgentOS turns your agents into a production API you can deploy anywhere.** It gives you a pre-built backend for your AI product, so all you need to do is add your agent logic, build your frontend and your product is live. +**Turn agents into production APIs.** AgentOS provides a pre-built backend for AI products. +Add agent logic, build the frontend, and ship. -You bring the agents and AgentOS turns it into a product: -- 50+ API endpoints with SSE-compatible streaming. Ready for production on day one. -- Sessions, memory, knowledge stored in your database. Your data stays in your system. -- Request-level isolation. No state bleed between users, agents and sessions. -- JWT-based RBAC with hierarchical scopes. Enterprise-ready from day one. -- Tracing built-in, stored in your database. No third-party egress. No vendor lock-in. -- Guardrails, human in the loop, learning machines. The full Agno framework. +## Create AgentOS -Here's an example of an AgentOS serving an Agent with memory, state, and MCP tools: + + + +Serve a single agent with memory and tools. Connect to the UI to monitor and debug. ```python agno_agent.py from agno.os import AgentOS @@ -37,15 +35,83 @@ if __name__ == "__main__": agent_os.serve(app="agno_agent:app", reload=True) ``` - + + +Pre-register components (agents, tools, workflows) to use them in the visual builder. + +```python agent_os_registry.py +from agno.agent.agent import Agent +from agno.db.postgres import PostgresDb +from agno.models.anthropic import Claude +from agno.models.openai import OpenAIChat +from agno.os import AgentOS +from agno.registry import Registry +from agno.tools.calculator import CalculatorTools +from agno.tools.duckduckgo import DuckDuckGoTools + +db = PostgresDb(db_url="postgresql+psycopg://ai:ai@localhost:5532/ai", id="postgres_db") + + +def sample_tool(): + return "Hello, world!" + + +registry = Registry( + name="Agno Registry", + tools=[DuckDuckGoTools(), sample_tool, CalculatorTools()], + models=[ + OpenAIChat(id="gpt-5-mini"), + OpenAIChat(id="gpt-5"), + Claude(id="claude-sonnet-4-5"), + ], + dbs=[db], +) + +agent = Agent( + id="registry-agent", + model=Claude(id="claude-sonnet-4-5"), + db=db, +) + +agent_os = AgentOS( + agents=[agent], + id="registry-agent-os", + registry=registry, + db=db, +) + +app = agent_os.get_app() + +if __name__ == "__main__": + agent_os.serve(app="agent_os_registry:app", reload=True) +``` + + + + +## Architecture + +AgentOS Architecture + +AgentOS Architecture -AgentOS runs on FastAPI. Add middleware, custom routes, background tasks, or any FastAPI feature without extra configuration. +AgentOS consists of two parts: - +1. **Runtime**: Runs agents built with the Agno SDK. +2. **User Interface**: Manages, monitors, and debugs the system. -## Control Plane +Runs as a container in your cloud. The UI connects directly from the browser. +No proxies. No data relays. -AgentOS comes with a built-in [control plane](https://os.agno.com) that connects directly to your runtime from your browser. Monitor sessions. Debug with full traces. Manage users and permissions. +FastAPI Compatible: Add routes, workers, and middleware without configuration.