Qarinah Reviews — Discover what people think of this product.

Qarinah

Cited project memory for coding agents

OssSelf-HostedFree product

Qarinah gives Codex, Claude Code, CLI tools, and MCP clients a compact, cited memory of the project in front of them. It stores decisions, code structure, tool outcomes, approvals, and source evidence locally, then retrieves only the context needed for the current task. In the published benchmark, 442,113 estimated repeated input-context tokens became 5,682 - 98.71% less repeated context.

This product has been submitted for review. Learn how to skip the line .

Get Notified

- supporters

What does Qarinah help with?

Qarinah is local, evidence-linked project memory for coding agents.

I built it because long-running projects kept paying to replay the same files, decisions, and tool history to every new agent session. Qarinah records permitted project events in a typed graph with deterministic Markdown and JSON views, then compiles a small cited context pack for the task at hand.

It works across Codex, Claude Code, CLI tools, and MCP clients. Memory packs can be exported and imported across supported environments, while every selected item points back to source evidence.

Published evidence: 442,113 estimated repeated input-context tokens became 5,682, with every required target covered in the top five. That is 98.71% less estimated repeated context.

Open source under Apache-2.0. Website: https://qarinah.io - Source: https://github.com/AjnasNB/qarinah

Qarinah Reviews — Discover what people think of this product.

Qarinah

What does Qarinah help with?

Tags