Vibe Coding Discoveragentao
View on GitHubLocal-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.
Agentao is a local-first, embeddable Python agent runtime: use it as a library, CLI, or ACP server. It ships permission modes, tool confirmation, MCP and skills, SQLite memory, sub-agents, and JSONL audit replay with no cloud dependency.
Use Cases
Embed a governed AI agent directly into a Python applicationRun an interactive terminal coding/automation agent with slash commandsServe the agent over ACP JSON-RPC for editor and IDE clientsConnect external tools via MCP stdio/HTTP/SSE serversEnforce tool confirmation and permission modes (read-only, workspace-write, full-access, plan)Keep SQLite-backed persistent user and project memory with auto recallReplay past sessions from JSONL recordings for auditOrchestrate built-in and custom sub-agents in foreground or backgroundInstall and auto-discover skill packs from GitHub reposSwitch between OpenAI, Anthropic, Gemini and other compatible providers at runtimeRun headless/CI batch jobs through the non-interactive prompt flagSandbox execution on macOS via sandbox-exec
Built With
- Language
- Python
- Frameworks
- MCP Python SDK · OpenAI Python SDK · Anthropic SDK · Pydantic · httpx · Rich · prompt_toolkit · Playwright · jieba · python-dotenv · VitePress
Tags
python · agent-runtime · local-first · mcp · acp · permissions · memory · audit-replay · cli · embeddable · openai-compatible · sub-agents · skills · sandbox · plan-mode · session-replay