Happy slice of mcp pizzaMCP.pizza

ax

MCP.Pizza Chef: google

This tool is built for developers who create AI agents that run long, multi-step jobs — the kind that can take minutes or hours to finish. If a connection drops or a process fails partway through, the agent picks up exactly where it left off instead of starting over. It runs on your own computer, on a server you manage, or on Kubernetes for larger deployments. Setup involves installing a command-line program and writing some configuration, so it suits developers comfortable with a bit of coding rather than someone wanting a simple point-and-click add-on.

Coding

Use This MCP server To

Resume a long AI task exactly where it stopped after a crash Keep several AI agents running at once without losing track of any Catch back up after losing connection, without restarting from scratch Run my own AI agent on a server I control See a step-by-step record of everything my AI agent did Recover an unfinished AI task instead of losing all its progress

README

Agent Executor (AX)

Warning

🚧 AX is in active early development.

We are actively refining our core, resumption protocols, and runtime specifications, which will introduce major breaking changes prior to a stable release.

Temporary Policy: We are temporarily pausing the acceptance of external Pull Requests while we stabilize the core architecture. We warmly encourage you to open Issues for feedback and feature requests instead.

AX, short for Agent Executor, is a distributed harness runtime. It dynamically provisions isolated environments from suspendable/resumable images to execute harnesses and agents. AX is designed for reliability, with native support for recovery and execution resumption, even in distributed setups.

Features

  • Distributed Runtime: Harnesses, skills, tools, and agents can execute in isolation
  • Resumption: Automatic recovery from failures or interruptions
  • Built-in Harnesses: Support for frontier harnesses and custom implementations
  • Portability: Runs anywhere, scales to small and large deployments
  • Customizability: Bring custom environment, MCP tools, skills, instructions, and more

Built-in consistency and resumability features:

  • Single-Writer Architecture: Single controller ensures consistent state management
  • Event Log: Durable execution state with automatic recovery
  • Advanced Resumption: Support for compute-layer actor resumption on compatible platforms

Demo

Demo

Watch our demo to see AX works when deployed on Agent Substrate.

Overview

ax FAQ

Do I need to know how to code to use this?
Yes — this is a developer tool. Setting it up means installing a command-line program (built with the Go programming language) and, for real deployments, working with Kubernetes. It is not something you install with a single click inside an app like Claude Desktop or ChatGPT.
Which AI assistant apps does this work with?
It is not a plug-in for consumer apps like Claude Desktop, ChatGPT, or Cursor. Instead, it is infrastructure a developer uses to build and run their own AI agents, which can in turn call models such as Google's Gemini, Anthropic's Claude, or OpenAI's models.
Do I need an account or a key to use it?
Yes. To use the built-in agent, you need a Gemini API key from Google AI Studio, or Google Cloud credentials if you are using Vertex AI instead.
Can I use this to keep an AI task running if my internet drops?
Yes — that is one of its main jobs. If a connection is lost partway through a task, you can reconnect and catch up on exactly what was missed, without restarting from scratch.
Can I use this to run more than one AI agent at a time?
Yes — it is built to manage many agents running independently at once, each in its own isolated space, which is useful if you are running several AI-driven jobs in parallel.
Is this ready for everyday, non-technical use?
Not yet. The project describes itself as being in active early development, with major changes possible before a stable release. It is aimed at developers building AI agent systems, not everyday users.
Where does this run?
It can run locally on your own machine, on any server you manage, or — for production use — on a Kubernetes cluster, which is the setup its creators recommend.
Is there a cost to use it?
The tool itself is free and open source, but you will pay for whatever AI model usage it triggers, such as Gemini API usage or your Google Cloud costs.