AWS Developers Open-Source Pizza Bot for AI Agents
A team of AWS developers has open-sourced Pizza Bot, an asynchronous, inbox-style interface that allows practitioners to run and monitor background AI agents without constant supervision.
AWS principal technologist Joseph Dolivo and solutions architect Igor Fil have released Pizza Bot, a self-hosted, Apache 2.0-licensed application. The tool acts as an email-style inbox for managing long-running AI agent workflows. Instead of requiring users to monitor active sessions, Pizza Bot treats tasks as asynchronous threads. Agents can execute scheduled or webhook-triggered work, delegate tasks to specialized workers, and pause to request human approval before taking consequential actions.
The client-server architecture is designed to keep data and agent states local. Pizza Bot stores threads, checkpoints, attachments, and logs in a user-controlled folder on the local machine, while storing credentials in the operating system's secret store. Data is only transmitted to the user's chosen model provider and explicitly enabled Model Context Protocol (MCP) servers. Dolivo and Fil emphasize that users retain complete control over what leaves their machine, though they must remain cautious about the permissions granted to enabled MCP servers.
The project began in April 2025 as a personal endeavor named JoeBot to automate CRM logging. Inspired by a LangChain article on ambient agents, the creators expanded the tool to accommodate non-technical users. While early reactions on platforms like Hacker News have been highly positive regarding the asynchronous inbox model, some developers have raised questions about how the community-driven project will replicate the ecosystem of integrations found in proprietary alternatives.
Because Pizza Bot is launched as a community project rather than an official AWS service, it does not come with AWS support or service-level agreements. For AI practitioners, this means they are solely responsible for deployment, backups, and updates. However, the tool offers a privacy-focused, local-first framework that allows developers to run complex, multi-agent workflows in the background without incurring the overhead of continuous session monitoring.
This is our own summary of reporting by InfoQ AI
