Polygraph is a meta-harness designed to enhance the capabilities of AI coding agents by providing them with a comprehensive understanding of entire codebases and enabling persistent session memory. It addresses the limitations of current AI agents that are often confined to a single repository or lack the ability to retain context across different interactions. By connecting all your repositories, both private and public, into a unified dependency graph without the need to move any code, Polygraph grants AI agents the visibility they require to operate more autonomously and effectively.
The core problem Polygraph solves is the inherent limitation of AI coding agents in understanding the broader context of a software project. Without a holistic view of the codebase, agents struggle to make decisions that are aware of dependencies, potential conflicts, or the overall architecture. This leads to fragmented work, repeated explanations, and a lack of continuity in AI-assisted development. Agentic amnesia, where AI agents forget previous interactions and context, is a significant bottleneck, hindering productivity and increasing the cognitive load on developers.
One of Polygraph's key features is its ability to create a unified dependency graph across all connected repositories. This graph acts as a synthetic monorepo, allowing AI agents to navigate and understand the relationships between different parts of the codebase. This visibility is crucial for tasks that span multiple services or modules, ensuring that changes made in one area are understood in the context of others.
Another significant capability is Polygraph's session memory persistence. Unlike many AI coding tools that treat each conversation as stateless, Polygraph ensures that session memory survives beyond the current interaction. This means developers and their AI agents can resume, reference, or build upon previous work without losing context. This feature is particularly valuable for long-term projects or when handing off tasks between team members.
Polygraph also facilitates cross-repository actions. For instance, if a change in a shared library affects multiple downstream repositories, Polygraph allows an AI agent to validate this change across all affected repos before any pull requests are opened. It can then manage these cross-repo pull requests and their associated continuous integration (CI) statuses cohesively, treating the entire change as a single unit.
Furthermore, Polygraph enables the seamless continuation of development sessions across different machines and even different agent frameworks. A session initiated by one developer on their machine can be picked up by another developer on a different machine, using a different AI agent, without any loss of context or state. This interoperability breaks down silos and promotes collaborative development.
The overall approach of Polygraph involves indexing repositories semantically to build a dependency graph, which agents use to find relevant code. It then checks out repositories locally and delegates sub-agents to work within them, ensuring access to actual code rather than just embeddings. This process allows for deep analysis and modification of code across multiple repositories.
The benefits for users include significantly improved autonomy for AI coding agents, reduced context switching for developers, and streamlined workflows for cross-repository changes. Developers can start working on complex features that span multiple repositories immediately, with all necessary context and setup managed by Polygraph. This leads to faster development cycles and fewer integration issues.
Concrete use cases for Polygraph include individual developers working on features that touch multiple repositories, where Polygraph sets up a single session, manages CI, and records all agent activity. For teams, Polygraph can validate changes across numerous downstream repositories before a single pull request is opened, managing all related PRs and CI statuses as a unified operation. It also aids in bug reproduction by allowing external parties to view sessions if all involved repositories are open-source.
Polygraph is designed for individual developers and teams working with multiple repositories. The product is free to use. While specific integrations are not detailed, it supports various agent frameworks, allowing session memory to persist across them. The underlying technology involves dependency graph traversal, semantic indexing, and local code checkout for agent execution.
In summary, Polygraph empowers AI coding agents with unprecedented cross-repository visibility and persistent session memory, transforming how developers collaborate with AI to build complex software systems more efficiently.
