Ref.

Ref builds planning and documentation tools for engineers working with AI coding agents. Ref Plans gives teams a shared workspace to research their codebase, review implementation decisions before code is written, and turn outlines into tasks with verification criteria. Teams can dispatch tasks to agents such as Claude Code, Cursor Background Agents, Codex and Devin, which report progress back to the plan. Ref Context lets MCP-compatible coding agents search public and private documentation, including private repositories and PDFs.

Founder Matt Dailey previously worked on applied AI at Figma and led Search & Indexing at Palantir. Ref’s documentation retrieval uses MCP session history to suppress repeated search results and select relevant sections when reading a page, returning roughly 5,000 tokens. This approach is designed to limit the documentation occupying an agent’s context window while allowing it to refine searches as it learns.

Ref serves individual developers, engineering teams and developers embedding documentation retrieval into their own agents. It sells subscriptions and additional usage credits, with credits shared across documentation search and AI-assisted planning; enterprise plans add SSO and dedicated support. In September 2025, the company reported thousands of weekly users and hundreds of subscribers, an early measure of adoption for its documentation-search offering.