Agentor

Agentor is an open source Python framework from Celesto AI for building, deploying and operating long running AI agents. It includes tool use, MCP support, the A2A agent to agent protocol and durable execution, and a working agent with tools can be created and served in a few lines of code. It is designed for developers who need agents that keep running reliably rather than single request chatbots. Durable runs save every step, so another process can finish a run after a crash or fork it into a new one. Agentor runs on its own agent engine with structured output and works with any OpenAI compatible provider, including OpenRouter, Groq, Together, vLLM, Ollama or a self hosted server. Calling serve on an agent enables the A2A protocol automatically, making it discoverable to other agents. The documentation covers tool calling and building MCP servers, tracing, skills, streaming, deployment and observability. The current stable release is version 0.1.0, installed with pip. The framework is free under the Apache 2.0 license. Celesto's cloud, which runs agents in sandboxes, uses prepaid usage pricing with no monthly platform fee: Micro at $0.03 per hour, Small at $0.04, Standard at $0.06, Memory at $0.08, Large at $0.16 and XL at $0.32 per hour. Credit is added in any amount above $2, and a trial gives a one time $5 runtime credit for 30 days without a card. The paid offer allows up to 10 running sandboxes, subject to host capacity, and stopped root disks are kept free for 30 days.