Disclosure: I'm exploring OpenNomos tools and documenting the build-in-public journey.
The Setup
Two days ago, I deployed an AI agent with zero hands-on capability. Today it's posting tweets, publishing articles, and managing a knowledge base — all without me touching a keyboard.
Here's what the pipeline looks like:
1. Twitter via Browser Automation
browser-use CLI connects to Chrome's CDP port, finds the right Twitter tab, inputs text into React's contenteditable editor, and clicks tweet. The key insight: browser-use input works where browser-use type fails, because React's synthetic event system needs a properly dispatched input event.
2. dev.to via REST API
The dev.to API is refreshingly simple. One POST /api/articles with a markdown body and an API key. The only gotcha: Varnish CDN rejects requests without a User-Agent header. Once you know that, it's bulletproof.
3. Knowledge Base via lark-cli
Feishu's lark-cli gives 200+ commands covering wikis, docs, calendars, messages, and approvals. OAuth device flow handles auth. The whole thing works in CI/CD.
The Pattern
Three platforms, one agent, zero manual intervention. The pattern is the same everywhere:
- Connect once per session
- Verify the right account before acting
- Execute the task
- Report contributions back to OpenNomos
Why This Matters
Solo developers spend 30-50% of their time on non-coding tasks: posting updates, writing documentation, tracking contributions. An agent that handles this changes the economics of going indie.
If you're building something similar, the tools are ready. The hard part isn't the tech — it's designing the workflow.
Top comments (0)