Applied reasoning, persistent agent memory, context engineering, and human-governed trust for AI coding agents — Claude Code, Claude Cowork, Codex, Cursor, Antigravity. Compatible with Obsidian. Free and open source. ARIA automates full knowledge lifecycle capture and human review, hook-enforced decision discipline and application at file edit, and codemap generation that traces full-stack flows for faster and smarter development, and more. Knowledge stored in markdown (Obsidian compatible).

AI is powerful, but with all its raw power, it frequently fails on context, memory, and enforcement. ARIA was built out of my own internal kit for optimizing Claude Code to execute across my projects precisely and efficiently, while avoiding repetition and mistakes that eat up time, effort, and token usage. It's also built around the premise that while the models are powerful, only you can validate what is valuable and should be actively applied. The free open-source plugin does three things: auto-captures session insights and intakes raw knowledge into staged backlogs, suggests contextually valuable knowledge to be promoted to active use, and enforces context surfacing + structured decision-making at session start and every file edit via pre/post hooks. It also captures snapshots automatically pre-compaction, so you don't lose valuable context. Everything is in skills and a local knowledge markdown base that works as an Obsidian vault. Fully accessible, usable across sessions and agents, and improves uniquely to your needs as you use it.  Try it out and let me know what you think!

We've been looking into this human in the loop thing - I think it's sth with a revolutionary potential.

@natalia_zak Thank you! Let me know how it works for you. There's a lot built-in but once you've run /extract a few times and done your first /audit-knowledge, it starts to learn much faster.

Just released v2.10.0 with major updates based on feedback plus optimizations for the new Opus 4.7 release.

Now available for Claude Code, Claude Cowork, Codex, Cursor, and Antigravity!

We've been looking into this human in the loop thing - I think it's sth with a revolutionary potential.

@natalia_zak Thank you! Let me know how it works for you. There's a lot built-in but once you've run /extract a few times and done your first /audit-knowledge, it starts to learn much faster.

Just released v2.10.0 with major updates based on feedback plus optimizations for the new Opus 4.7 release.

Now available for Claude Code, Claude Cowork, Codex, Cursor, and Antigravity!

ARIA Knowledge

Active knowledge capture and decision enforcement for Claude

Active knowledge capture and decision enforcement for Claude