A new open source project called gcontext aims to solve one of the most frustrating problems with AI-assisted coding: context loss when you return to a project after time away. The tool, which launched on Hacker News this week, lets AI agents maintain persistent notes about their work—stored as plain markdown in git—so they never lose track of integrations, active tasks, or where things stand. The core premise is straightforward but powerful. When you're deep into debugging a complex feature or building out an API integration, you might step away for a few days. Come back, type "continue" into your AI assistant, and watch it scramble to reconstruct what it was doing. gcontext flips this by having the agent keep its own running documentation—automatically updating notes about stack integrations, current work in flight, and project-specific context that persists across sessions. What makes this particularly interesting from a developer tooling perspective is its broad compatibility. The system works with Claude Code (Anthropic's coding assistant), Cursor, and OpenAI's Codex. This multi-platform approach means you're not locked into one vendor's proprietary memory system—you get persistent context regardless of which AI tool you're using in any given session. The notes live directly in your project's git repository as markdown files. For development teams, this is a significant advantage: the context isn't trapped in some external service or database. It's version-controlled alongside your code, accessible to any team member who clones the repo, and survives across machine changes and environment rebuilds. When a new developer joins a project, they can see exactly what the previous AI-assisted sessions were working on—valuable context that usually vanishes the moment you close a chat window.
Open Source First, Cloud Version Coming
The project is open source with an MIT license, meaning developers can inspect, modify, and contribute to the codebase. A cloud-hosted version is planned for future release, which will presumably offer a more turnkey experience for teams that don't want to manage their own CLI setup—but the core functionality remains available as self-hosted software today.
When Persistent Context Actually Matters
The use cases practically write themselves once you start thinking about them. Complex debugging sessions that span multiple days become manageable when your AI assistant remembers every hypothesis it tested and path it ruled out. Feature development that gets interrupted by priorities can resume seamlessly. Onboarding onto legacy projects becomes faster when you have an AI that already understands the architectural decisions encoded in the codebase—not just what the code does, but why certain choices were made.
The Bottom Line
gcontext addresses a real gap in the current AI coding assistant experience, and doing it through standard git workflows makes it immediately adoptable for teams already living in version control. At 4 points on Hacker News with zero comments, this is early-stage—but that's exactly when these kinds of utility tools prove their value. If you're working with AI agents on anything beyond quick one-off tasks, persistent context isn't a luxury—it's a necessity.