agentmaker
MIT Python agent framework with memory, hybrid retrieval, RAG, MCP, and governance in one import.
agentmaker is an open-source Python framework (pip package 0.2.0, Python 3.12+) that ships a unified agent loop for chat, ReAct, plan-and-solve, and reflection, plus built-in hybrid retrieval, Mem0-style long-term memory, RAG, MCP tool loading, guardrails, human-in-the-loop approve/resume, run limits, and tracing to JSONL, SQLite, or OpenTelemetry—with local SQLite defaults and swappable backends.
Official resources
Selection advice
Quick comparison
agentmaker packages agent loops, retrieval, memory, RAG, and governance in one MIT core. Pydantic AI emphasizes typed structured outputs, while Agno focuses on a lightweight path from SDK toward hosted AgentOS.
| agentmaker | Pydantic AI | Agno | |
|---|---|---|---|
| Best for | Python teams wanting agents, memory, and hybrid retrieval without stitching many packages | Schema-validated tools and structured outputs on Pydantic models | Lightweight agent SDK with a clear hosted AgentOS path |
| Batteries included | Hybrid retrieval, memory, RAG, MCP extras, guardrails, HITL, tracing | Type-safe agents; long-term memory and vector stores are mostly bring-your-own | Agent SDK plus optional cloud runtime; memory/knowledge via framework features |
| Tradeoff | Younger project (created 2026-07-08) with a smaller ecosystem than LangGraph-class stacks | Lean typing core vs fewer built-in retrieval/memory services | Broader production footprint vs less of a single-import full stack |
Best for
- single-import Python agent stacks
- local-first hybrid retrieval and memory
- MCP tool integration with guardrails and HITL
- teams that want swappable SQLite-default backends
Not ideal for
- TypeScript-first teams
- teams that need a large battle-tested graph orchestration ecosystem
- products that only need a thin typed agent wrapper without retrieval or memory
Core concepts
Minimal implementation shape
pip install agentmaker, create `Agent("assistant", LLMClient("deepseek"))`, call `agent.run(...)` for multi-turn chat with memory, then add tools, MCP, or RAG extras as needed.