AI Engineer - Tools

Arcade•San Francisco, CA

About The Position

Arcade is building the MCP runtime that gives AI agents the power to securely take action inside enterprise systems. This involves handling authorization, governance, and trust when agents access customer data, execute workflows, or make changes on behalf of a user. Arcade provides a tools catalog, actions platform, and governance model to enable AI agents to perform real actions on real systems, already in use within Fortune 100 companies. This role is crucial for developing the agentic tools that allow AI models to interact with and manipulate external systems, moving beyond simple chat capabilities.

Requirements

4+ years of software engineering experience shipping production code.
Strong Python and Typescript skills.
Experience building and consuming APIs, including REST, auth flows, pagination, rate limits, and handling real-world integration complexities.
A testing mindset with the ability to write tests that catch real failures.
Comfort with ambiguity in an early-stage team environment with a charter that will expand and decisions made with incomplete data.
An insatiable desire to ship.

Nice To Haves

LLM application experience, including prompting, retrieval, tool use, or agent design.
Familiarity with the MCP (Model Context Protocol) ecosystem or similar agent-tool protocols; experience filing an issue against the spec is a plus.
Experience building evaluations or measurement systems for ML/AI behavior.
Prior experience at an API platform, integrations-heavy product, or developer tools company.
Open-source contributions.
Experience at an early-stage startup, and enjoyment of that environment.

Responsibilities

Build the agents used internally to build, maintain, and scale toolkits and skills.
Build new toolkits by taking a vendor's API and transforming it into a high-quality set of tools that agents can call reliably.
Maintain and improve the existing catalog by fixing breakages, sharpening tool descriptions and response shapes, and ensuring high quality as upstream APIs change.
Write tests and evaluations that prove a tool does what it claims, against real API behavior and across different models.
Tune tool descriptions, tool-use behavior, and how an agent calls the built tools, treating LLMs and agents as a first-class part of the job.
Contribute to the shared patterns and tooling that accelerate the building of new toolkits.
Engage with and push the MCP ecosystem, where standards for agent tools are still developing.