Product
How NORDON Works
Think of NORDON as an intelligent notebook for your AI coding assistant. It captures, scores, and retrieves the important stuff from every coding session using 40+ signal detectors and a 6-signal ranking algorithm.
The Pipeline
Capture. Score. Retrieve.
Imagine you had an assistant who watched every coding session, scored what matters, and briefed your AI before every new conversation. That's NORDON.
Watch & Capture
While you code with your AI assistant, NORDON quietly watches what happens. 40+ signal detectors scan for architecture decisions, performance insights, security concerns, and dependency changes.
Score & Classify
Not everything is worth remembering. A weighted ML scoring algorithm rates each event from 0 to 1. Only events above the 0.4 threshold become memories. NLP pattern matching classifies them into 7 structured types.
Rank & Recall
When you start a new session, the 6-signal ranking formula finds the most relevant memories. Recency decays with a 14-day half-life. Acceptance feedback adjusts scores. Your AI picks up exactly where you left off.
Built in Rust
No Python. No heavy ML frameworks.
NORDON's entire ML engine is written in Rust. 16 modules. Zero cloud dependencies. Retrieval in under 50ms. Scoring in under 5ms per event. 265 automated tests. 80,000+ lines of code. A single binary that runs on macOS, Windows, and Linux with no runtime dependencies.
Explore the ML EngineMemory Types
Seven kinds of things NORDON remembers
Your project isn't just code. It's decisions, patterns, mistakes, and rules. NORDON understands the difference and stores each type in a way your AI can use.
Facts
Things that are true about your project — like which database you use, what version of React you're on, or where your API lives.
Decisions
Choices your team made and why — like why you picked Postgres instead of MongoDB, or why you went with REST instead of GraphQL.
Procedures
Step-by-step instructions for how to do things — like deploying to production, running database migrations, or setting up a new developer's machine.
Failures
Things that went wrong and what you learned — so your AI assistant never makes the same mistake twice.
Patterns
The way your team likes to do things — naming conventions, file structures, coding styles. The unwritten rules of your codebase.
Constraints
Hard rules that can't be broken — security requirements, performance limits, browser compatibility, or regulatory rules.
Context Snapshots
What's happening right now — the current sprint goal, an ongoing refactor, a migration in progress. The stuff that changes week to week.
Smart Retrieval
6-signal ranking for precise recall
NORDON doesn't dump everything into your AI's context. A multi-signal ranking formula picks the right memories based on what you're working on right now.
Keyword Match
Memories about the same topic or files you're currently working on get the highest weight.
Importance Score
ML-powered importance scoring ensures critical decisions and failures rise to the top.
Recency Decay
Recent events are more relevant. 14-day half-life decay ensures old context fades gracefully.
Acceptance Feedback
Memories your AI actually used and you approved get a higher score in future sessions.
Scope Match
Memories from the right branch, project, and part of the codebase are prioritized.
Frequency
Memories that come up repeatedly across sessions are likely important to surface.
Privacy
Everything stays on your computer
NORDON runs entirely on your machine. Your code, your decisions, your memories — they never leave your computer unless you choose to share them.
Runs on Your Machine
The entire memory system runs locally. No cloud servers, no API calls to external services.
SQLite Database
All your memories live in a single file on your computer. You can inspect it, back it up, or delete it anytime.
Share When Ready
Want to share memories with your team? You can — with end-to-end encryption. But it's always your choice.
Zero Tracking
No telemetry, no analytics, no usage tracking. We don't know what you're building and we don't want to.
Integrations
Works with the tools you already use
Claude Code
Plugs right in via hooks. Memories are automatically captured and injected — no setup needed beyond install.
VS Code
See memory hints inline, capture new memories with a command, and browse everything in a side panel.
Git
NORDON understands branches. Feature branch context stays separate from main. Memories follow your code.
CLI
Full control from your terminal. Search, inspect, add, or delete any memory. Export and import as needed.
Visualization
Knowledge Graph
See your project's knowledge as an interactive force-directed graph. Memories become nodes, relationships become edges. Decisions connect to constraints, failures link to fixes, patterns cluster around file groups.
Interactive Exploration
Click, drag, and zoom through your project's knowledge. Filter by memory type, time range, or importance score. See how decisions, patterns, and failures are connected.
Relationship Discovery
NORDON automatically detects relationships between memories -- a decision that created a constraint, a failure that led to a pattern, a procedure that depends on a fact.
Cluster Detection
Memories that are frequently accessed together or relate to the same files naturally cluster. See which parts of your codebase have the richest context.
Conversation ML
Conversation Intelligence
NORDON doesn't just watch tool results -- it understands your conversations. When you discuss architecture decisions, debug a tricky bug, or explain a constraint to your AI, those insights become structured memories automatically.
Decision Extraction
When you tell your AI 'let's use REST instead of GraphQL', NORDON extracts that as a DECISION memory with the reasoning and context.
Constraint Detection
Phrases like 'we can't use X because Y' or 'this must always be Z' get captured as CONSTRAINT memories that prevent future violations.
Failure Learning
When you debug an issue and explain what went wrong, the root cause, symptoms, and fix all get structured into a FAILURE memory.
Pattern Recognition
Discussions about naming conventions, code structure, or team workflows become PATTERN memories that keep your AI aligned with team practices.
Quality
Smart Suggestions & Session Summaries
NORDON doesn't just store memories -- it helps you improve them. Every memory gets a quality grade, and actionable suggestions tell you how to make your knowledge base better.
A-F Quality Grades
Every memory gets a letter grade based on completeness, specificity, and actionability. See which memories need improvement.
Actionable Tips
NORDON suggests specific improvements -- add context, link related files, specify the reasoning behind a decision.
Repo Health Score
An overall score for your project's memory health. Track improvement over time as you refine your knowledge base.
Session Summaries
When a session ends, NORDON auto-generates a summary of what happened -- decisions made, bugs fixed, patterns established.
Get Started
Give your AI assistant a memory
Install the CLI and your AI coding assistant will never forget what matters again.
npm install -g @sodasoft/nordon-cli