Hi Product Hunt, I’m Pradeep, the maker of ContextHarbor. AI coding agents can read our repositories, but understanding the code is only part of understanding a project. Why did we choose this architecture? Which requirement makes that seemingly redundant check necessary? What did we try before, and why did we reject it? Why did John make this change ? ( when John is on leave) Those answers often live in specifications, meeting notes, runbooks, or someone’s memory. We end up reconstructing them whenever we start a new session or switch tools. I built ContextHarbor to make that knowledge available inside the development workflow. With ContextHarbor, you can: - Bring project documents together and retrieve relevant passages with sources. - Preserve decisions, constraints, and lessons as durable project memory. - Connect Claude Code, Codex, Cursor, and other MCP clients to shared project knowledge. - Organize knowledge by project, with access controls for your team. - Deploy on infrastructure you control and configure your embedding and LLM providers. For example, before changing a scheduled job, an agent can retrieve the architecture decision explaining its timing constraint. That gives it useful evidence before proposing a change. ContextHarbor is a commercial product, with evaluation access available on request. I’d love your feedback: What piece of project knowledge do you keep having to explain to your AI coding agent?

Context-Harbor

Persistent project context for AI coding agents

Persistent project context for AI coding agents

ContextHarbor

Previous Context-Harbor Launches

Previous Context-Harbor Launches