Machina¶
The open-source Python framework for building AI agents specialized in industrial maintenance.
Machina is to industrial maintenance what LangChain is to general-purpose LLM applications. Build AI agents that talk to your CMMS, read your equipment manuals, monitor your sensors, and help your technicians — all in a few lines of Python.
Install¶
Optional extras for specific integrations:
pip install machina-ai[litellm] # OpenAI / Anthropic / Ollama / 100+ providers via LiteLLM
pip install machina-ai[cmms-rest] # GenericCmmsConnector REST mode (httpx)
pip install machina-ai[docs-rag] # DocumentStoreConnector with ChromaDB vector search
pip install machina-ai[telegram] # TelegramConnector for technician chat
pip install machina-ai[all] # Everything
Where to start¶
-
:material-rocket-launch: Quickstart
Go from
pip installto a working Maintenance Knowledge Agent in 30 minutes, using sample CMMS data and manuals — no external accounts required. -
:material-book-open-page-variant: Domain Model Reference
Complete API reference for
Asset,WorkOrder,FailureMode,SparePart,Alarm,MaintenancePlan, andPlant. Aligned with ISO 14224. -
:material-sitemap: Architecture
The five-layer architecture: connectors, domain model, agent runtime, LLM abstraction, and observability. How data flows end-to-end.
-
:material-puzzle: Custom Connectors
Build your own connector by implementing a simple
BaseConnectorProtocol. No ABCs, no decorators — just a Python class that exposescapabilities. -
:material-file-cog: YAML Configuration
Define your agent entirely in
machina.yaml— connectors, LLM, channels, plant. Load withAgent.from_config("machina.yaml"). Zero Python needed.
Why Machina?¶
Building an AI maintenance agent today means writing custom connectors for SAP PM, IBM Maximo, or whatever CMMS your plant uses. It means defining domain concepts like assets, work orders, and failure modes from scratch. It means handling OPC-UA subscriptions, Modbus registers, and MQTT topics. And it means engineering prompts that understand maintenance — all before writing a single line of business logic.
That takes months. Machina makes it take minutes. Pre-built connectors for
industrial systems, a rich domain model aligned with ISO 14224, and maintenance-aware
AI — so you can go from pip install to a working agent in under 30 minutes.
License¶
Machina is licensed under Apache 2.0. Contributions welcome — see CONTRIBUTING.md.