Software Engineer, Agent Harness

Cursor; San Francisco, New York (CA, NY); On-site; USD 180,000 - 230,000; Full time

Job summary

Cursor in San Francisco seeks a Software Engineer for the Agent Harness team. You’ll build core agent behavior and capabilities powering experiences across Cursor products, including agent orchestration, tools, guardrails, and model tuning—designed as a platform SpaceXAI engineers can extend.

Your work will define the default Cursor agent experience ("Auto"), where users rely on smart model choices under real-world constraints, balancing performance and cost with trust.

Qualifications

• Built complex agentic products or infrastructure.
• Strong software engineering fundamentals and production systems.
• Experience with models and agent behaviors, and staying updated on AI research.

Responsibilities

• Build core agent behavior and capabilities powering Cursor products.
• Develop agent orchestration, tools, guardrails, and model tuning as a platform for SpaceXAI integrations.
• Shape the default Cursor agent experience and smart model choices under real-world constraints.

Skills

Agentic products

Software engineering

Model behavior

Model eval datasets

Research awareness

Product tool development

Cross‑team collaboration

Job description

Our mission is to automate coding. The first step in our journey is to build the best tool for professional programmers, using a combination of inventive research, design, and engineering. Our organization is very flat, and our team is small and talent dense. We particularly like people who are truth-seeking, passionate, and creative. We enjoy spirited debate, crazy ideas, and shipping code.

About the Role

As a Software Engineer on the Agent Harness team at SpaceXAI, you’ll build the core agent behavior and capabilities that power agent experiences across Cursor products. This spans agent orchestration, tools, guardrails, and model behavior tuning—designed as a platform that SpaceXAI engineers can extend and depend on.

Your work will also define the default Cursor agent experience (”Auto”), where users trust us to make the right model choices under real-world constraints.

Example projects include:

• Advancing the agent harness (agent loop, tools, prompts, execution environment, capabilities) used across Cursor products. This would shape and empower experiences like multi-agent coordination.

• Improving the default Cursor agent experience (the “Auto” model setting): making smart model choices based on user needs, model capabilities, and cost appetite.

• Shaping the default Cursor agent’s personality.

• Launching new models on Cursor end-to-end — partnering with foundational model providers and our research team on integration, evaluation, and rollout.

You may be a fit if

• You have built complex agentic products or infrastructure.

• You have strong software engineering fundamentals and enjoy shipping production systems.

• You have taste and strong opinions on model and agent behaviors. You stay up-to-date and informed on emerging research and industry trends.

• You make good calls in the gray area: weighing data, user experience, and product taste when there isn’t a single “right” answer.

• Experience building eval datasets or frameworks to understand and improve the performance of AI systems is a plus.

• Experience building products or tools used by other engineers is a plus.

#LI-DNI