OpenComputer launches as open-source Devin alternative
DiggerHQ has released OpenComputer, an open-source coding agent designed to be faster than Devin and capable of self-hosting in private clouds.
Source: Hacker News — front pageDiggerHQ has introduced OpenComputer, an open-source coding agent designed to serve as a high-speed alternative to Devin. By providing a self-hostable stack, the project aims to give development teams more control over their AI-driven coding workflows.
What is OpenComputer and how does it work?
According to the project documentation, OpenComputer is an AI coding agent that integrates directly with communication tools like Slack. Users can trigger tasks by mentioning @Computer in a Slack thread, after which the agent operates on its own computer to complete the assignment. Once the work is finished, the agent submits the results as a pull request.
The architecture is designed for flexibility and security. The entire stack can be hosted on a single Virtual Machine (VM), which the developers recommend configuring with outbound-only network access to enhance security. This VM runs a gateway named Calvin, which acts as a bridge between Slack events and various AI models. Depending on the user's specific configuration, the agent can utilize Claude Code, Codex, or OpenCode. To ensure process isolation, each session is managed using rootless Podman, with state being persisted to a git-backed store after being written to the local disk.
How fast is OpenComputer compared to other agents?
One of the primary value propositions of OpenComputer is its speed. The developers claim the tool is built to match the responsiveness of running Claude Code locally on a laptop.
In benchmarks provided by the project, OpenComputer reportedly opens pull requests 3.3x faster than Devin. Specifically, the documentation notes that all four pull requests in their test case were opened with a green CI status in 56 seconds, compared to 185 seconds for Devin. Furthermore, the tool is cited as being 2.6x faster than the average of the four most popular cloud-based agents, achieving a first push in 24 seconds versus a 62-second average.
Can OpenComputer be self-hosted?
Yes, self-hosting is a core feature of the project. Because OpenComputer is open-source from the compute layer upward, teams can run the entire infrastructure within their own cloud accounts. This provides an alternative to purely managed services, allowing for greater data sovereignty and customization.
While the stack is optimized to run on a single VM, the developers utilize Cloudflare Workers and Durable Objects for other parts of the stack. They have also indicated that support for other platforms, such as Fly, is planned for the future. For those who prefer not to manage infrastructure, OpenComputer also offers a cloud-based version that can be set up in minutes.
What are the licensing and cost options?
OpenComputer is released under the Apache 2.0 license, making it accessible for various types of development projects. For users opting for the cloud version, the project offers a three-day free trial that includes 600 minutes of work. Additionally, the service allows users to "bring your own" subscriptions for models like ChatGPT or Codex, rather than being locked into a single proprietary billing model.
How does this relate to modern AI coding workflows?
As AI agents move from simple autocomplete tools to autonomous engineers capable of managing entire pull requests, the tension between convenience and control grows. Open-source alternatives like OpenComputer allow developers to audit the agent's behavior and keep the compute within their own security perimeters. This is particularly relevant for enterprises that may be hesitant to send proprietary codebases to closed-source third-party platforms.
For developers looking to experiment with integrated AI coding environments, Grafy provides a specialized workspace. While OpenComputer focuses on agentic automation via Slack and self-hosted VMs, Grafy Code offers a dedicated AI coding studio where users can build and run code directly. Exploring different model capabilities is also possible through Grafy Chat, which allows for side-by-side comparisons of various AI models.
Frequently asked questions
What license does OpenComputer use?
OpenComputer is released under the Apache 2.0 license. This open-source licensing allows developers and companies to use, modify, and distribute the software according to the terms of the Apache license, facilitating widespread adoption and community contribution.
How can a team use OpenComputer in Slack?
A team can use OpenComputer by mentioning @Computer within a Slack thread alongside a specific coding task. The agent then performs the work on its own computer and eventually delivers the completed task as a pull request for the team to review.
Is it possible to run OpenComputer on private infrastructure?
Yes, OpenComputer is designed to be self-hostable. Because the project is open-source from the compute layer up, teams can run the entire stack on their own Virtual Machines or within their own cloud accounts to maintain control over their data and security.
Source: Hacker News
Source: Hacker News — front page — Show HN: OpenComputer – an open source alternative to Devin


