Run code agents and a personal assistant on a custom agent factory
Developer at a consultancy uses Codex, Claude and their own Workforce agent harness for daily back-office software development, plus a personal assistant called Clara. ·
- What they did
- Dave Sanders, who works at a consultancy building back-office systems for customers, uses the Codex and Claude desktop apps directly. He is moving toward his own agent factory, Workforce, which runs general and code agents in containers for some containment. Workforce agents build and deploy Workforce itself. He also uses it to run Clara, a personal assistant he is building. Models mostly go through OpenRouter.
- How it went
- He has agents coding 8-10 hours a day plus personal projects after work, and just stays under the $200 Claude plan limits. He expects that to get uncomfortable as he adds agents.
- Worth knowing
- Subscription limits are the constraint: he pays for a $200 Claude plan and a $100 Codex plan, and more agents will strain the Claude cap.
Try it yourself with Claude
Act as my back-office software developer on [repo path]. Take this task: [describe task, e.g. add an export feature to the invoicing tool], work on a new branch, run the tests, and leave a summary of what changed and anything you're unsure about. You're done when the tests pass and the branch is ready for me to review. Don't merge, push, or touch production config without showing me the diff first.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: HN · Nov 5, 2026
Related use cases
Build a C compiler with a team of parallel agents
Anthropic's engineering write-up on tasking Opus 4.6 agent teams with building a C compiler.
Compile and filter apartment listings
A renter used Claude and ChatGPT to compile property-management sites and filter listings; results were decent but included some out-of-parameter listings, and he found a place through a friend.
Screen interview code submissions with a Codex review agent
A commenter uses Codex with a targeted agents.md for front-line code reviews, rejecting submissions with too many files changed or added.
Review Linux kernel patches
Google engineers launched Sashiko, an agentic AI code reviewer for the Linux kernel.
Custom agent
Build a clinic app with booking and trackers
A doctor with no coding experience used Replit to build a clinic app in 5 days with booking integration, prescription requests, health trackers and an AI symptom checker.
Build specs through automated multi-agent verification and review loops
A developer writes a spec, then lets an implementation engine build it with review rounds from different agents and models, triaging issues and looping until the code meets spec.
Custom agent
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.