Polygraph

Polygraph is a meta-harness built by the Nx team that extends AI coding agents with two capabilities they fundamentally lack: visibility across every repository boundary, and memory that persists across sessions. It is currently free during early access, with a 30-day notice promised before any pricing changes.

What It Is

Polygraph sits above the agent layer — it does not replace Claude Code, Codex, or OpenCode, but augments them. It connects all repositories a user has access to (private and public OSS) into a unified dependency graph without moving any code, giving agents the ability to read and write across every repo and orchestrate PRs and CI as a single coordinated change. The docs describe it as delivering "synthetic monorepo" ergonomics: monorepo-style cross-repo awareness with no migration required.

The Two Walls Polygraph Addresses

The docs articulate the core problem as two constraints that cap agent autonomy:

• Space: An agent is confined to one repo at a time and cannot see how a change fits the wider system or write to multiple repos simultaneously.
• Time: Every session starts from a blank slate, forcing a human to re-carry context across sessions, machines, and teammates.

Polygraph lifts both by building a cross-repo dependency graph and recording every agent session for resumption or reference.

Core Features

• Automatic repo discovery: Polygraph indexes every repo the user has access to and builds a graph of relationships via packages and APIs, so agents plan cross-repo work without being told where to look.
• Cross-repo PR and CI coordination: Open, link, track, and verify related PRs across repos from a single session.
• OSS repo integration: Bring public OSS repos into a session for upstream debugging, repros, and integration validation.
• Memory and resumability: Resume a session the next day, reference it from a new session, or hand it off to a teammate — repos, branches, PRs, and logs transfer with it.
• Session history: Reference any past session by its description, PRs, and traces to fix bugs without digging through history.

Supported Agents and Integrations

Polygraph plugs into an existing stack without replacing it. The homepage lists integrations with:

• Codex (OpenAI)
• Claude Code (Anthropic)
• GitHub
• Open Code

Existing repository permissions are respected, so no new access model needs to be established.

Use Cases

The docs outline several concrete scenarios Polygraph is designed for:

• Shipping a feature that spans a backend and its frontend, or a contract across microservices
• Validating a design system or library change against every consumer, then PRing the clean ones
• Patching a CVE across every affected repo with a full audit trail
• Cross-repo impact analysis and pattern investigation
• Maintaining a single session across an entire sprint to avoid re-explaining context at every handoff
• OSS developer workflows: triaging contributor repros and validating fixes against consuming repos

Current Status

Polygraph is in early access and is currently free. The product page states "Full product, zero cost. 30 day notice before any pricing changes." It is built by the Nx team, the organization behind the Nx monorepo build system, and is positioned as a natural extension of that ecosystem's philosophy applied to the agentic coding era.