agentao

Local-first, governed AI agent runtime for Python — embed it in your app, or run it as a CLI or ACP server. Permissions, MCP, memory and audit replay built in.

What is agentao?

Local-first, governed AI agent runtime for Python — embed it in your app, or run it as a CLI or ACP server. Permissions, MCP, memory and audit replay built in.

Overview of agentao

What it is

Agentao is a local-first, governed AI agent runtime for Python. It can be embedded in an application, run as a CLI, or used as an ACP server. The runtime includes permissions, MCP, memory, and audit replay as built-in features. Its name combines “Agent” (capability) and “Tao” (governance), reflecting its design around constraint, connectivity, and observability.

Key features

  • Governance: tool confirmation, four permission modes (read-only, workspace-write, full-access, plan), and macOS sandbox support.
  • Memory: SQLite-backed persistent memory with user and project scopes, automatic recall, and Chinese segmentation via jieba.
  • Protocols: MCP (stdio, Streamable HTTP, SSE) for tools and ACP (stdio JSON-RPC) for full agents.
  • Skills: auto-discovered from a skills/ directory, installable from GitHub via agentao skill install owner/repo[:path][@ref].
  • Sub-agents: built-in codebase-investigator and generalist, plus custom agents defined in .agentao/agents/.md.
  • Replay & Output: JSONL session recordings under .agentao/replays/, markdown toggle, and /copy for the last reply.
  • Embedding: Agentao(...) constructor with events() stream, active_permissions(), capability injection, and ACP Pydantic schemas.

Use cases

  • Embedding a governed agent into a Python application with full control over inputs and permissions.
  • Running an interactive agent CLI with slash commands, plan mode, and memory inspection.
  • Using Agentao as an ACP server to manage agent sessions remotely.
  • Developing and testing custom skills, tools, and sub-agents in a local-first environment.

Getting started

bashpip install agentao``````pythonfrom pathlib import Pathfrom agentao import Agentaofrom agentao.llm import LLMClientfrom agentao.transport import NullTransportwith Agentao( working_directory=Path('/tmp/agent-run-1'), llm_client=LLMClient( api_key='sk-...', base_url='https://api.openai.com/v1', model='gpt-5.4', ), transport=NullTransport(),) as agent: reply = agent.chat('Summarize today's logs.') print(reply)For CLI usage, install with the [cli] extra and set the required environment variables (OPENAI_API_KEY, OPENAI_BASE_URL, OPENAI_MODEL). See the repository README for full details.

License and pricing

This project is licensed under the MIT License. It is open source with no stated pricing.

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/jin-boagentao.svg)](https://yunodb.com/tools/jin-boagentao?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/jin-boagentao?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/jin-boagentao.svg" alt="Featured on YunoDB" height="20"></a>