You have probably heard about AI coding agents that charge $20 or $30 per task. OpenHands just shipped a version that runs on your own machine for the cost of a few cents in API calls.
What's happening
OpenHands is an open-source autonomous coding agent: give it a plain-English task, and it writes and runs code, checks the output, and keeps iterating until it finishes. The v1.17.0 release, out September 9, ships a cleaner task-outcome dashboard and smarter automation runs that actually tell you what the agent accomplished and what it skipped.
The project is MIT-licensed with more than 74,000 GitHub stars. Benchmarks put it at roughly 68 percent task completion on the standard SWE-bench software-engineering suite, which matches or beats several commercial agents that charge per task. People in the self-hosted AI community are regularly using it for site maintenance, data cleanup, and small internal tools that would have needed a hired developer a year ago.
What makes this interesting for a small shop: you bring your own API key (a few cents per session with a mid-tier model), or point it at a local Ollama model for fully offline, zero-cost-per-query operation. Either way, your data stays on your machine.
Try this this week
- Visit github.com/OpenHands/OpenHands and scroll to the Quick Start section.
- If you are comfortable with a terminal, one Docker command spins it up locally. If not, the hosted demo at openhands.dev lets you test it before installing anything.
- Pick one task you have been postponing because it felt like it needed a developer: a CSS tweak on your site, a script to rename a folder of photos by date, a formula to calculate something in your spreadsheet. Write it in plain English and let the agent attempt it.
- For tasks involving client data or internal files, connect it to a local Ollama model. Nothing leaves your machine, and there is no per-query cost.
- Always review what the agent did before deploying anything to a live site or touching real client data. Low-stakes tests first.
The bigger picture
One of the most meaningful shifts in AI right now is the gap closing between "small org with some tech curiosity" and "small org that can actually ship working tools." An open-source coding agent that handles real tasks for near-zero cost is a concrete example of that. If you want to figure out which repetitive tasks in your shop are worth automating, that conversation is a good place to start with build9.