Let an AI agent build a new project feature by feature on a recurring schedule
Keep a side project moving forward automatically while you are away from the keyboard
Fall back to a free local model when your cloud API key hits a rate limit
| vicsanity623/cronbuild | 100/dotfiles | abdellaouiamir/palworld-server | |
|---|---|---|---|
| Stars | 1 | 1 | 1 |
| Language | Shell | Shell | Shell |
| Last pushed | — | 2016-11-18 | — |
| Maintenance | — | Dormant | — |
| Setup difficulty | moderate | easy | moderate |
| Complexity | 4/5 | 1/5 | 3/5 |
| Audience | developer | developer | ops devops |
Figures from each repo's GitHub metadata at analysis time.
Requires the opencode CLI, GitHub CLI, Git, and Perl installed plus an API key for OpenRouter or Gemini.
CRONBUILD is a script that tries to build and improve your software project automatically, on a recurring schedule, without you sitting there writing code. You point it at a project, run its setup once to add an API key, and after that a scheduled job wakes it up repeatedly, each time reading a file called MEMORY.md to see what has already been built and what should come next. On each run, CRONBUILD hands that context to an AI coding agent called opencode, which is expected to read the project's files, implement the next feature, create a Git branch, commit and push the change, and open a pull request on GitHub. Another step then merges that pull request and appends a note to MEMORY.md describing what was built and what to build next. If a run fails, CRONBUILD resets the workspace, pulls the latest main branch, and retries with the previous failure included as context, up to ten attempts per day before giving up on that cycle. To avoid ever getting completely stuck, CRONBUILD tries up to three different AI models in order: a paid or free model through OpenRouter first, Google's Gemini as a fallback if that is rate limited, and a local Ollama model as a last resort that costs nothing and has no rate limit. The README also describes a lock file mechanism so that two scheduled runs cannot overlap, and a memory compaction step that trims MEMORY.md once it grows past a few days so the file does not get too long to fit in the AI's context. The README lists example day-by-day plans for building things like a browser game or a SaaS backend over one to two weeks, suggesting the tool is meant for greenfield projects as much as existing ones. Setup requires the opencode CLI, the GitHub CLI, Git, and Perl, with Ollama as an optional extra for offline use. The project is a single bash script plus a setup script, and it is released under the MIT license.
A scheduled script that uses AI coding agents to autonomously build, commit, and merge features into your project over time.
Mainly Shell. The stack also includes Shell, opencode, GitHub CLI.
Use freely for any purpose, including commercial use, as long as you keep the copyright notice.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.