Polygraph
Polygraph is a meta-harness that gives coding agents cross-repo visibility and session memory. It does not replace Claude Code, Codex, or OpenCode. Instead, it indexes every repo you can access, maps how they connect through packages and APIs, and records agent sessions so work survives handoffs, restarts, and tool switches.
Agents normally stall at two limits: one repo at a time, and no memory between sessions. Polygraph addresses both by turning distributed repos into a synthetic monorepo your agent can read and write across, while coordinating PRs and CI from a single session. Past sessions stay searchable with descriptions, branches, PRs, and traces attached.
It is built for engineering teams running multi-repo products who want one prompt to land changes everywhere without re-explaining context. Integrations include Codex, Claude Code, GitHub, and Open Code. Polygraph is built by the Nx team and is free during early access.
Indexes private and OSS repos into a dependency graph automatically
Agents read and write across every connected repo from one session
Open, link, and track PRs and CI across repos in one place
Session history with descriptions, PRs, and traces you can resume or reference
Hand off sessions between agents without losing repo or branch context
MCP tools for spawning agents, pushing branches, and fetching CI logs
Integrates with Codex, Claude Code, GitHub, and Open Code
Connects repos into a dependency graph so agents plan cross-repo work without manual context.
Session memory survives restarts and agent handoffs, reducing repeated explanations.
Free during early access with 30 days notice before any pricing change.
Pricing after early access is not published yet.
Focused on multi-repo engineering workflows, not general-purpose chat or content tasks.
Is Polygraph a coding agent?
No. Polygraph is not an agent. It is a meta-harness that wraps the agents you already use, such as Claude Code and Codex, adding cross-repo visibility, PR coordination, and session memory they lack on their own.
How does Polygraph handle multiple repositories?
Polygraph indexes every repo you have access to and builds a graph of how they relate via packages and APIs. Your agent uses that graph to plan and execute work across repo boundaries without being told where to look, similar to working inside a synthetic monorepo.
Can I resume a Polygraph session later or hand it to a teammate?
Yes. Polygraph records every agent session with its description, repos, branches, PRs, and traces. You can resume a session, reference it from a new one, or hand it to a teammate or different agent without re-explaining context.
Which coding agents work with Polygraph?
Polygraph integrates with Codex, Claude Code, GitHub, and Open Code according to its homepage. It also exposes MCP tools and a CLI for session, repo, agent, and CI management documented at trypolygraph.com/docs.
Is Polygraph free?
Yes. Polygraph is free during early access with the full product at zero cost. The site states Polygraph will give 30 days notice before any pricing changes.
Who built Polygraph?
Polygraph is built by the Nx team, the company behind the Nx monorepo build system. The product homepage and docs are hosted at trypolygraph.com.

