Software Engineer, Agent Harness
Builds core agent harness platform for agent orchestration, tools, guardrails, and model tuning powering Cursor products. Requires strong software engineering skills, experience with complex agentic systems, and judgment for model choices under constraints.
About the job
Responsibilities
- Build core agent behavior and capabilities powering agent experiences across Cursor products, including agent orchestration, tools, guardrails, and model behavior tuning.
- Design as a platform that Cursor engineers can extend and depend on.
- Define default Cursor agent experience (“Auto”), making smart model choices based on user needs, capabilities, and cost.
- Advance agent harness (agent loop, tools, prompts, execution environment, capabilities).
- Improve “Auto” model setting.
- Launch new models end-to-end, partnering with model providers and research team on integration, evaluation, and rollout.
Requirements
- Built complex agentic products or infrastructure with taste and strong opinions on model and agent behaviors.
- Strong software engineering fundamentals and experience shipping production systems.
- Good judgment in gray areas, weighing data, user experience, and product taste.
Nice-to-Haves
- Experience building products or tools used by other engineers.
Skills
Python, Agent Orchestration, LLMs, Prompt Engineering, Model Evaluation, TypeScript, React, Kubernetes, Infrastructure, Production Systems
Similar jobs
Fullstack Engineering jobsSenior full-stack engineer responsible for taking ambiguous 0-to-1 product concepts from idea through prototype, validation, and launch. The role emphasizes strong frontend development, backend and systems architecture, product judgment, and cross-functional collaboration.
Software Engineer building end-to-end financial data platform features including ingestion, transformation, automation engines, reporting, monitoring, and AI-driven workflows. Requires CS degree and 4+ years software engineering experience; onsite in SF or NYC.
Product engineer building the modern AI-first platform for accounting and finance. Tackles foundational data automation, transformation, and reporting challenges in a complex domain. Requires 5+ years software engineering experience and a bachelor's in CS, Math or Philosophy.
Build scalable full-stack web applications across frontend and backend layers using React, TypeScript, Python, GraphQL, and FastAPI. The role requires at least three years of full-stack development experience and offers hybrid work in San Francisco.
Product Engineer responsible for building Bolter’s AI agent runtime, workspace, generated applications, and production reliability systems. The role requires strong full-stack TypeScript experience, product judgment, rapid prototyping ability, and depth in AI systems, real-time infrastructure, or developer tooling.