← Sticker Wall AGENT · DETAIL

Using Devon for Open Source Project Backlog Issue Cleanup Monthly Subscription Service, Earning 8,000 RMB/Month

Workflow: Every morning, pull the Open Issue list of the target repository, filter by labels marked as bug or good first issue, so

AGENT

Key Fields

FIELD STAMPS

IndustryAI / LLM

RegionGlobal

ScaleSME

ChannelOnline

🔧 Workflow

Every morning, pull the Open Issue list of the target repository, filter by labels marked as bug or good first issue, sort by difficulty and reproduction clarity, and feed three to five into Devon. The Agent automatically clones the repository, reads the code, locates the root cause, writes the fix, runs tests, and finally initiates a PR. Manually review each PR's code quality, boundary conditions, and test coverage. Once confirmed correct, push it to the maintainers for review. The deliverables include a weekly settled list of merged PRs, failed issue post-mortem records, and backlog clearance progress reports.

🛠 Setup Requirements

Requires proficiency in Git and GitHub operations, the ability to read common backend or frontend code for manual review, and basic command-line familiarity without requiring deep algorithmic expertise. Run Devon locally or on a cloud server using Docker, connect to OpenAI or Anthropic model APIs, and prepare a clean GitHub account, personal order-taking homepage, and historical PR portfolio. Initial environment setup and running the first demo issue takes about two to three days. It is recommended to start within a familiar language ecosystem to lower the review difficulty.

🧰 Toolchain

🔧 Devon Open Source Agent
🔧 GitHub
🔧 OpenAI or Anthropic API
🔧 Docker

💰 Revenue

Charge a monthly maintenance fee agreed upon with open-source project maintainers or commercial companies, ranging from 2,000 to 4,000 RMB per project per month, including a commitment to a fixed number of merged PRs. Serving two to three projects simultaneously during the startup phase yields about 4,000 to 8,000 RMB per month. Once a reputation is established, prices can be raised for commercial companies relying on the open-source project, or charged additionally based on the number of merged PRs, with a single-person upper limit of about 12,000 RMB per month.

💸 Cost

Model APIs are billed by usage, with each issue fix consuming a few RMB in token costs, totaling around 200 to 600 RMB per month; cloud server or local machine electricity costs about 100 RMB per month; basic tools like GitHub and Docker are free. Overall monthly costs can be kept under 800 RMB, with the primary initial cost being the time investment for manual review.

⏱ Time Investment

2 to 3 hours per day, mainly used for selecting suitable issues, reviewing PRs submitted by the Agent, responding to maintainers' revision comments in the PR discussion area, and a weekly progress synchronization.

🚀 Getting Started

The first step is to find three to five open-source projects on GitHub with which you have hands-on experience, noticeable issue backlogs, and still-active maintainers. Use the Agent to help submit two high-quality fix PRs for free to build credibility. After securing the first batch of merge records, organize them into case study screenshots, and pitch the monthly backlog clearance proposal in project discussion areas, maintainer emails, and developer communities, starting with a pilot single project at a monthly fee of 2,000 RMB.

🔑 Keys to Success

✅ Manual review serves as a safety net to guarantee PR quality; it is better to submit fewer PRs than bad code that ruins your reputation
✅ Prioritize issues with small scope, full test coverage, and clear reproduction steps to increase the merge rate
✅ Only take on projects whose tech stack you understand; review capability is the core barrier of this business
✅ Organize every merged PR into a public portfolio, using verifiable records to drive the next client signing

⚠️ 风险

⚠️ Code generated by the Agent may introduce hidden bugs. Once merged into the main branch, triggering production issues will cause maintainers to lose trust and terminate cooperation
⚠️ After major tech companies open source similar capabilities for free, such as GitHub Copilot Agent mode and Codex, the pricing for monthly manual services may be continuously driven down
⚠️ Some open-source communities have an aversion to AI-generated code or require explicit labeling. Failure to comply with community norms may result in being blacklisted, harming personal GitHub reputation

📌 Real Cases

📌 Devon was launched by Proxify, capable of autonomously analyzing GitHub Issues, writing fix codes, and submitting PRs. Its release in 2025 sparked widespread discussion among developers regarding automated code maintenance and monetization paths
📌 Similar paths have been validated: in Devin's public demo, AI completed an 8-month legacy code migration plan in 8 days, showing that end-to-end autonomous fixing is indeed feasible in backlog cleanup scenarios, with coding efficiency reported to be several times that of humans
📌 Alibaba Cloud Developer Community reported practices already existing in the GitHub ecosystem: CI failures trigger Codex to automatically analyze and directly submit fix PRs, and the github-sre-agent project goes a step further by implementing autonomous monitoring, root cause analysis, and remediation, demonstrating that the workflow for automated issue cleanup is operational in real-world scenarios

Sources