> Markdown version of [/jobs/ext/4313088-founding-harness-engineer-ai-agents](https://www.wearedevelopers.com/jobs/ext/4313088-founding-harness-engineer-ai-agents). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# Founding Harness Engineer - AI Agents

- **Company:** ONE STOP COLLECTIBLE CORP
- **Location:** United States (Remote available)
- **Experience:** Expert
- **Contract:** Permanent contract
- **Skills:** Application Programming Interfaces (APIs), Python (Programming Language), TypeScript, Large Language Models, Backend, Langfuse — LLM Observability and Analytics Platform, LangSmith
- **Published:** October 11, 2026
- **Apply:** https://startup.jobs/founding-harness-engineer-ai-agents-us-only-bolder-apps-10417672

## About the Role

* You've shipped tool-using LLM agents to real users in production. Demos and prompt wrappers don't count

 * Good instincts for agent initiative, and the test results to back them up
 * Hands-on experience building eval and regression suites for LLM behavior
 * You understand safe tool execution, prompt injection, and multi-tenant data risk
 * Solid backend skills in Python or TypeScript (async jobs, retries, state)
 * You take a fuzzy problem from prototype to production and can explain the tradeoffs along the way
 * Excellent written and verbal communication; you can explain a tradeoff to a founder or a user without a whiteboard

## Description

We move fast. We take ownership. We work with AI, not against it. And we expect everyone to bring ideas, not wait for instructions.

There are no daily checklists, no micromanagement, and no corporate politics. You get autonomy, trust, and a small team that's always ready to help you grow. Here, impact matters more than titles, and curiosity matters more than seniority.

If you want to level up fast and see your work in real users' hands - welcome aboard.

Responsibilities

 * Build the agent runtime: context, tool choice, approval waits, and knowing when to act, ask, or stay quiet
 * Write and keep tuning the system prompts. Proactive is great. Spamming the team channel is not
 * Own the eval and regression harness. Prompt changes are code, so they get tested like code
 * Make long-running work survive interruptions, retries, and the occasional flaky API
 * Put the guardrails in: approval gates for risky actions and strict separation between customers' data
 * Work directly with the founder and early users, and turn what you hear into shipped behavior, * MCP, or integrations with tools like Slack, Gmail, Notion, or Linear
 * Eval and tracing tools such as Braintrust, Langfuse, or LangSmith
 * You've been an early engineer at a startup before and still liked it