Dalexor MI: The MCP Server That Gives Your AI a Permanent Memory of Your Codebase

If you've ever spent a morning deep in a complex problem with an AI coding assistant — only to open a new file and have it ask "can you paste the code you're talking about?" — you've experienced what Dalexor's founder calls the Goldfish Effect.

That frustrating amnesia is exactly what Dalexor MI is built to solve.

What Is Dalexor MI?

Dalexor MI is an MCP (Model Context Protocol) server that gives AI assistants like Claude and Cursor a persistent, searchable memory of your entire codebase history. Instead of re-explaining context every session, you install it once and let it build a living timeline of your project in the background.

It works with Claude Desktop, Cursor, and Windsurf out of the box, and can be installed in under two minutes:

pip install dalexor
dalexor init

The Problem It's Solving

Most AI coding tools use basic file search to pull in relevant snippets. The result is an AI that sees isolated fragments — but never the story of why your code is the way it is.

Dalexor's founder, Dragos Marincas, described the motivation on DEV Community: he didn't want to build another "chat with your files" wrapper. He wanted something that felt like a permanent brain for his project — a tool that tracks not just what changed, but why.

How It Works

Dalexor runs a lightweight CLI watcher (dx watch) in the background. Every time you save a file, it evaluates whether the change is meaningful. Trivial edits like formatting tweaks or whitespace changes are ignored. Substantive changes — a new authentication flow, a refactored data model — get captured as a Logic Snapshot, complete with an AI-written plain-English summary.

These snapshots are encrypted locally before they ever leave your machine, then stored permanently in your private cloud account and indexed for instant natural language search.

The three-step flow is:

1. Connect — Install the CLI and run dalexor init
2. Monitor — The watcher silently captures meaningful changes in real time
3. Search — Ask questions in plain English directly inside Claude or Cursor

What You Can Actually Ask It

Once Dalexor has been watching your project for a while, you can ask your AI assistant things like:

• "Why did we choose PostgreSQL?" → "Switched from MongoDB for ACID compliance..."
• "When did we add rate limiting?" → "Added 2024-12-15 after DDoS incident..."
• "Find the definition of UserSession" → Returns the snapshot where it was introduced
• "Show me all API changes" → Returns a filtered list of every relevant snapshot

That's a fundamentally different experience from paste-and-explain.

Key Features

AI Summarization — Every meaningful change gets an automatic plain-English note. Passive documentation that writes itself.

Conflict Prediction — Before you commit, Dalexor checks if anyone on your team has recently touched the same files or functions, warning you about likely merge conflicts before they happen.

Dependency Mapping — See exactly what else in the codebase depends on a function before you change it. Know the full impact of an edit before you make it.

Multi-Project Support — Run as many projects as you want under one account and search across all of them at once.

Visual Timeline Dashboard — Browse your project's full change history visually, scrolling through who changed what and when.

REST API Access — Hook Dalexor into CI/CD pipelines, Slack workflows, or any internal tooling via its REST API.

Security Architecture

Dalexor is built on a local-first security model. Raw code is filtered and encrypted on your machine before anything is transmitted. Encryption keys belong to you — Dalexor never holds them. Each organization's data is fully isolated at the database level, meaning no other Dalexor user can access your project history.

The dual-mode architecture — a local monitor handling encryption and IDE context, paired with a cloud engine for AI summarization and fast search — means the most sensitive processing never leaves your device.

Real-World Use Cases

Onboarding new developers — New team members can ask the project memory why architectural decisions were made, instead of interrupting senior engineers.

Debugging production issues — When something breaks, Dalexor surfaces when the relevant code changed and why, without combing through a year of git history.

Compliance and auditing — For regulated industries, Dalexor maintains a permanent, searchable, tamper-evident record of every significant change, who made it, and when.

The Backstory

Dalexor's founder started with an AI security monitoring tool that failed because he'd never validated whether a market existed for it. Rather than quit, he reframed the lesson: "I don't consider it a failure. I paid a price to learn new stuff." That pivot led to Dalexor MI — a product born directly from the pain of building with AI assistants that kept forgetting everything.

Getting Started

Dalexor MI has a free tier and requires no credit card to start. After installing and running dalexor init, you add the MCP bridge to your Claude Desktop config and immediately gain access to 11 tools including find_definition, trace_dependency, and predict_conflicts.

If you've ever lost momentum because your AI forgot what you just built, this is worth trying.

→ dalexor.com