Intelligent Memory
for AI Coding Agents
NORDON's Rust-powered ML engine scores, ranks, and learns from every coding session. Your AI assistant gets smarter with every conversation.
The Intelligence Engine
A Scoring Engine That Actually Learns
Signal Detection
Our engine scans every event for 40+ signals across 4 categories: architecture decisions, performance insights, security concerns, and dependency changes.
Importance Scoring
A weighted scoring algorithm rates each event from 0 to 1. User-flagged items get 0.4 boost. Decisions score 0.3. Failures score 0.25. Only events above 0.4 become memories.
Smart Classification
Events are automatically classified into 7 memory types using NLP pattern matching: decisions, failures, procedures, constraints, patterns, facts, and context snapshots.
Retrieval Ranking
When your AI needs context, a 6-signal ranking formula finds the most relevant memories. Keyword match (30%), importance (20%), recency with 14-day half-life decay (15%), acceptance feedback (15%), scope match (10%), frequency (10%).
Feedback Learning
Every time you approve or reject a memory, the engine adjusts. Acceptance rates influence future scoring. The system gets more accurate over time.
Performance
Built in Rust. Blazing Fast.
NORDON's core engine is written in Rust for maximum performance and reliability.
No Python. No heavy ML frameworks. Just clean, fast, reliable Rust code that runs anywhere.
Comparison
What Sets NORDON Apart
| Feature | Chat History | Simple Memory | NORDON |
|---|---|---|---|
| Structured types | × | × | 7 types |
| Importance scoring | × | × | ML-powered |
| Retrieval ranking | × | Basic | 6-signal |
| Branch-aware | × | × | ✓ |
| Dependency tracking | × | × | Auto-detect |
| Pattern detection | × | × | ✓ |
| Drift alerts | × | × | ✓ |
| Secret redaction | × | × | 31 patterns |
| Knowledge graph | × | × | Interactive |
| Conversation extraction | × | × | ML-powered |
| Quality scoring | × | × | A-F grades |
| Local-first | × | Maybe | Always |
| Multi-agent | × | × | 7 agents |
Features
Everything your AI assistant needs to remember
Knowledge Graph
Visualize your AI's memory as an interactive graph. See decisions, patterns, and failures as connected nodes. Understand how your project's knowledge is structured at a glance.
Conversation Intelligence
NORDON learns from your conversations, not just tool results. When you discuss architecture or debug a bug, those insights become memories automatically.
7 Structured Memory Types
Not just chat logs. NORDON stores decisions, patterns, failures, procedures, constraints, facts, and context snapshots as organized knowledge your AI can actually use.
6-Signal Smart Retrieval
A multi-signal ranking formula weighs keyword match, importance, recency decay, acceptance feedback, scope, and frequency to surface the right context at the right time.
Branch-Aware Context
Working on a feature branch? NORDON keeps that context separate. No mixing up main branch knowledge with experimental work.
Memory Quality Scoring
Every memory gets an A-F quality grade. Smart suggestions help you improve low-quality memories. Your repo gets an overall health score so you know where to focus.
Private by Default
Everything stays on your computer. No cloud. No data sharing. 31 secret patterns auto-redacted. 28 blocked file types. Just your machine, your memories.
Session Summaries
When a session ends, NORDON auto-generates a summary of what happened -- decisions made, bugs fixed, patterns established. Your next session starts with full context.
Interactive Review
Approve or reject memories right from the CLI. NORDON learns from your feedback -- accepted memories get boosted, rejected ones get suppressed.
Trusted by Developers
Developers shipping faster with NORDON
“I stopped repeating my project's architecture every session. NORDON just knows. It's like having a senior engineer's context permanently available.”
— Solo developer, open-source maintainer“Our deployment failures dropped 40% after we started capturing failure memories. The ML scoring surfaces exactly the right warnings at the right time.”
— Engineering lead, Series B startup“Onboarding time went from 2 weeks to 3 days with shared team memory. New hires' AI assistants already know our architecture decisions and gotchas.”
— CTO, 50-person engineering teamCompatibility
Works with your stack
Get Started
Start with the free plan. Your AI will thank you.
Install the CLI and give your coding AI the intelligence it deserves. Two minutes to set up. Free forever on personal projects.
$ npm install -g @sodasoft/nordon-cli$ nordon init$ nordond &$ nordon link-repo$ nordon doctor