Polygraph

Polygraph is a meta-harness for AI coding agents. It gives them something they typically lack: visibility across multiple repositories and a memory that persists across sessions. Instead of an agent being limited to a single repo, Polygraph indexes every repo you have access to—private and open-source—and maps how they relate through packages and APIs. This means you can prompt your agent once and have it work across your entire codebase, without manually telling it where to look.

How does it work? Polygraph automatically discovers repos, builds a dependency graph, and lets agents plan cross-repo changes. You can test and preview downstream effects before committing. It centralizes pull requests and CI status in one interface, respecting existing permissions. A standout feature is session history: you can reference any past session, and Polygraph brings back its description, PRs, and traces, so you never have to re-explain context. You can also hand off a session across any agent (Codex, Claude Code, Open Code) and everything—repos, branches, logs—comes along.

Polygraph is built for engineering teams working with AI agents on complex, multi-repo codebases. It's especially useful for teams that need to automate cross-repo changes, share agent context among teammates, or debug issues that span repos. During early access, it's free. The tagline "see everything, forget nothing" captures its core promise.