Pylon Releases Open-Source Daemon That Turns Webhooks and Cron Schedules Into Sandboxed AI Coding Agent Runs

Pylon, a new open-source platform for orchestrating AI coding agents, is now available as a self-hosted daemon that turns external events into sandboxed agent runs. Aimed at engineering teams that want to integrate AI-driven code analysis into their workflows without surrendering control of source code or infrastructure, Pylon spins up coding agents, including Claude Code and OpenCode, inside isolated Docker containers, each scoped to a configurable workspace.

The platform connects to existing developer tooling through webhooks and cron schedules. A new Sentry error event, a GitHub pull request, or a nightly cron tick can each trigger a pipeline that runs an agent against the relevant codebase. Configurable workspace modes: full git clone, lightweight git worktree, local directory mount, or code-less execution; let teams match isolation level to the sensitivity of the task. Approval workflows on high-volume triggers ensure agent runs only happen when an operator confirms, while Slack and Telegram notifications keep human reviewers in the loop.

Pylon ships with built-in templates for the most common automation patterns: Sentry error triage, GitHub pull-request review, and scheduled security or quality audits. A web-based dashboard called Nexus provides interactive management, including job history, live logs, and the ability to attach to, kill, or retry running agents. Configuration is hot-reloadable, so operators can adjust pipelines without restarting the daemon.

The project is fully open source and runs on Linux and macOS across both amd64 and arm64 architectures. The only runtime requirement is Docker. Documentation, installation guides, and the configuration reference are available at https://docs.pylon.to, and the source code and issue tracker are hosted at https://github.com/pylonto/pylon.

Pylon is positioned for development teams that want the productivity benefits of agentic AI inside their own perimeter, running on their hardware, in their network, against their code, rather than handing source over to a third-party SaaS. By making each agent run sandboxed, observable, and approvable, the project aims to bring the operational discipline of CI/CD to AI-assisted code workflows.

Pylon is available now at https://www.pylon.to

Pylon
Bojan Stefanovic
411 Walnut Street # 22887
Green Cove Springs, FL 32043-3443

Press contact: Bojan Stefanovic
Email: press@pylon.to
Phone: ‪(646) 783-9308‬
Web: https://www.pylon.to

Pylon is an open-source software project providing a self-hosted daemon for running sandboxed AI coding agents triggered by external events. The platform was built to give engineering teams a way to incorporate agentic AI into their development workflows, code review, error triage, scheduled audits, while keeping source code, agent runtime, and orchestration entirely within their own infrastructure. Pylon supports Claude Code and OpenCode as agent backends, runs each job inside an isolated Docker container, and integrates with common developer tools including Sentry, GitHub, Slack, and Telegram. The project is distributed under an open-source license and maintained by Bojan Stefanovic.

This release was published on openPR.