DeepSeek Open-Sources an Agent That Answers Your GitHub Issues
In short
- DeepSeek put an open-source coding agent on GitHub. It watches one repo and opens a pull request when an approved person asks for work.
- MIT license, about 3.3k stars within days. DeepSeek says it is a research prototype, not an official product, and not production-ready.
- The money angle: small repetitive tasks are what you pay a freelancer for. This agent targets exactly those, and the repo lists no price.
DeepSeek just gave away an agent that answers your GitHub requests by itself. You write the task. It writes the code and replies in the same thread, usually with a pull request. How much does it cost? The repo states no price. The license is MIT.
What happened
On October 31, DeepSeek published dsh-superagent, as reported by ShiftDelete. It is a proactive coding agent for one GitHub repository, packaged for DeepSeek Harness. It listens to GitHub, and optionally to Feishu group chats. When an allowed person asks for work, it opens a work item, does the job in its own session and directory, and answers where the request came from.
There is also a web panel that shows only the items that need a human. The stated idea: take as little human attention as possible and give as much value as possible. Several tasks can run in parallel sessions.
Now the honest part. The repo calls it a research prototype for reference only. It is not an official DeepSeek product and not production-ready. The page shows no release date, and the repo has a single commit. Treat it as a very interesting blueprint, not a finished employee.
How to use it
- Start on a test repo. Not your live shop. Pick a copy with low stakes.
- Let your own coding agent set it up. The repo’s SETUP.md is written for that: tell your agent to read it and set the project up with you. Manual steps are in docs/006-deployment.md.
- Limit who can ask. Only allowed people trigger work. Keep that list short.
- Give it boring tasks. Rename things, fix small bugs, update docs, add tests.
- Review every pull request. You are the gate. Nothing merges without you.
Do the math: the repo gives no price, so I build this on the cost it replaces. A freelancer fixing small issues at €30 each, 20 issues a month, is €600 a month and €7,200 a year. If the agent handles even a quarter of them, that is €150 a month you keep. What you pay for it is the model usage you connect, and that depends on your setup. The numbers are my example.
What it means
The pattern is the news. A request lands in a normal place, GitHub or a chat. An agent picks it up, works alone, comes back with a result. A human only steps in when needed. That is how small teams will run software in a year.
The weak point is trust. An agent that acts on requests needs strict permissions. After last week’s agent-security stories on this site, I would not point it at anything that can touch money or customer data.
Ivan’s Take
Free and MIT is the right way to ship this, and I like the idea a lot. But it is a prototype, so I would test it on a throwaway repo and read every line it proposes. The day it is stable, it replaces the cheapest layer of freelance dev work.
Sources: DeepSeek · ShiftDelete
If it doesn’t save you time or money, you won’t see it here
- Facts, not hypeEvery price and number is checked against the official source
- Always the real priceYou see what it costs before you click
- I say when it’s weakEven when it’s OpenAI or Google
