Builders who wrote their own agent loop, harness or editor around Claude, Codex and open models, and what each one adds over the stock tools. 18 matching use cases below, each with a public source.
Omnara is an open-source, self-hostable alternative to Claude Managed Agents and OpenAI's Agents API. You create an agent with an API call; Omnara runs the agent loop, gives it sandboxes to run commands in and keeps runs going.
See pricing page1 dayAdvanced
Lugh Xtudio is a desktop video studio where Claude Code, Codex or Antigravity does the editing on a real timeline and you can still adjust anything by hand. It is aimed at creators and faceless channels.
$17/moWeeksIntermediate
Vercel released durable-harness, which wraps Pi, OpenCode, Codex and Claude Code so a run that is killed mid-task can pick up where it stopped instead of forgetting everything. It works locally, on Vercel or on your own infrastructure.
$17/mo1 dayAdvanced
Vercel extended the durability system behind its Workflows product to any agent harness, including Pi and OpenCode, as open source. It runs locally, on Vercel or on your own servers, so agent runs can be resumed after failures.
$17/mo1 dayAdvanced
The founders of Relay, a local-first agent supervisor, pre-registered a 5-test suite with hidden judges and ran their tool, Claude Code, Prime Agent and Hermes through the same token meter on Sonnet 5.5 and Haiku 5.5, then published the losses along with the wins.
$17/mo1 dayAdvanced
He had Claude Code build a code editor shaped around how he works, and talks to his agents through Full-Duplex Code, an open-source voice layer for Claude Code and Codex.
$17/mo100 minIntermediate
He built an open-source server where you name the harness (Claude Code, Codex, Pi or about 20 more) and send a request, with each run sandboxed. Use it for one-off jobs, build agents on top, or write loops instead of prompts.
$17/mo1 hourAdvanced
A developer built CloudHarness MCP, a whole coding agent harness inside one MCP server: pi agents run inside it and carry out tasks delegated from ChatGPT, skills install from any GitHub repo, and credentials are injected without showing up in chat.
See pricing page1 dayAdvanced
Based on a 2026 Harness Learning paper, a developer built a coding-agent harness that studies its own failed runs and rewrites itself. He kept the model (Claude Sonnet 5.5) and tasks fixed and only let the harness change.
$2 in / $10 out per 1M tokensDaysAdvanced
The follow-up to the viral parents'-TV post shares the exact prompt, the launcher source and the gotchas the builder hit, so anyone can point their own coding agent at an Android TV or Google TV and get the same clean home screen.
$17/mo1 hourBeginner
A developer built his own agent harness workspace on top of Claude Code, Codex, OpenCode and Grok Build. All sessions share one SQLite database, work streams are Discord-style channels, and each agent has its own persona, memory, skills and scheduled jobs.
$17/moWeeksAdvanced
Built Rhun, a small cross-platform editor in assembly with Vim mode, terminal, git diffs and panels for Claude Code or Codex.
$17/moweeksAdvanced
Shipped HarnessRouter: Unified interface for agent harnesses and shared it as a Show HN.
$17/mo1-2 daysAdvanced
Built Pu.sh, a full coding-agent harness in about 400 lines of shell.
$17/mo1 hourIntermediate
Built an open-source agent that the author reports topped TerminalBench on Gemini-3-flash-preview.
See pricing page1-2 weeksAdvanced
Two Neovim fans wanted an agent harness they could customize as deeply as their editor, so they built Ito, a Lua library for terminal UIs. Its browser playground ships an ito.md guide you paste into your LLM along with what you want to make.
See pricing page1-2 hoursIntermediate
Built a LangGraph workflow plus C++ harness so agents can run, benchmark and Nsight-profile CUDA kernels.
See pricing page1 weekAdvanced
Built Peerd, an AI agent harness that runs entirely in the browser.
See pricing page1-2 hoursIntermediate