https://github.com/microtronx/mxlore
Self-hosted MCP server for AI coding assistants. Persistent knowledge base with 39 tools, 11 skills, institutional memory.
https://github.com/microtronx/mxlore
ai-tools claude-code delphi developer-tools knowledge-management mariadb mcp mcp-server
Last synced: 2 months ago
JSON representation
Self-hosted MCP server for AI coding assistants. Persistent knowledge base with 39 tools, 11 skills, institutional memory.
- Host: GitHub
- URL: https://github.com/microtronx/mxlore
- Owner: MicrotronX
- License: other
- Created: 2026-04-06T18:36:17.000Z (3 months ago)
- Default Branch: main
- Last Pushed: 2026-04-07T15:33:27.000Z (3 months ago)
- Last Synced: 2026-04-07T15:39:20.800Z (3 months ago)
- Topics: ai-tools, claude-code, delphi, developer-tools, knowledge-management, mariadb, mcp, mcp-server
- Language: Pascal
- Homepage: https://www.mxlore.dev
- Size: 336 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE.txt
Awesome Lists containing this project
README
# mxLore
Self-hosted MCP server for AI-assisted software development. Stores architectural decisions, specs, plans, findings, and lessons learned — accessed via [Model Context Protocol](https://modelcontextprotocol.io/) tools from AI coding assistants.
> **New here?** Check out the **[product overview at mxlore.dev](https://www.mxlore.dev)** first — see what mxLore does, how it compares, and why your AI assistant needs a memory.
## Features
- **39 MCP Tools** — search, create, update, recall knowledge across projects
- **Multi-Project** — one server, many codebases, cross-project search
- **Skill Evolution** — tracks which AI checker rules actually help, auto-tunes
- **Institutional Memory** — lessons learned with recall, gate levels, graph-based knowledge links
- **Multi-Agent** — agents communicate across projects via message bus
- **Admin UI** — web dashboard for developers, keys, projects, intelligence metrics
- **AI Batch** — automatic summaries, tagging, stale detection at boot time
## Quick Start (5 minutes)
### Prerequisites
- Windows x64
- MariaDB 10.6+ (recommended: 11.x)
- Claude Code CLI (for AI features)
### 1. Install MariaDB
Download from [mariadb.org](https://mariadb.org/download/) and create an empty database:
```sql
CREATE DATABASE mxai_knowledge CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci;
```
### 2. Configure
Copy and edit the config template:
```bash
copy mxLoreMCP.ini.example mxLoreMCP.ini
```
Set your database password in `mxLoreMCP.ini`:
```ini
[Database]
Password=your_db_password
```
That's the only required change. MariaDB is auto-detected from your installation.
### 3. Start
```bash
mxLoreMCP.exe
```
The server will:
- Auto-detect your MariaDB installation (registry, common paths)
- Create the database schema automatically on first boot
- Start the MCP server on port 8080
- Start the Admin UI on port 8081
### 4. First Setup
Open `http://localhost:8081` in your browser. The Admin UI is open without login on first start (no developers yet). Create your first developer and API key.
### 5. Connect Claude Code
First, download the setup skill:
```bash
# Windows (PowerShell)
mkdir "$env:USERPROFILE\.claude\skills\mxSetup" -Force
Invoke-WebRequest -Uri "https://raw.githubusercontent.com/MicrotronX/mxLore-skills/main/mxSetup/SKILL.md" -OutFile "$env:USERPROFILE\.claude\skills\mxSetup\SKILL.md"
# Linux / macOS
mkdir -p ~/.claude/skills/mxSetup
curl -o ~/.claude/skills/mxSetup/SKILL.md https://raw.githubusercontent.com/MicrotronX/mxLore-skills/main/mxSetup/SKILL.md
```
Then in Claude Code, run:
```
/mxSetup
```
Enter your server URL (`http://localhost:8080/mcp`) and API key when prompted. mxSetup installs all remaining skills, hooks, proxy, and configures the MCP connection automatically.
## Network Security
| Scenario | Setup |
|----------|-------|
| **Localhost** (single developer) | HTTP is fine. Default: `BindAddress=127.0.0.1` |
| **LAN** (team, internal network) | Change `BindAddress=0.0.0.0` in INI. HTTP acceptable if network is trusted. |
| **WAN / Cloud** (public internet) | **HTTPS required.** Use a reverse proxy (IIS, nginx, Apache) with TLS certificates. |
## Password Security
The INI file supports two formats for database and API passwords:
- **Plain text** (default, simplest): `Password=mypassword`
- **XOR obfuscation** (optional): `PasswordEnc=` — generate with `mxLoreMCP.exe --encrypt "mypassword"`
> **Note:** XOR obfuscation prevents casual reading of passwords in the INI file. It is NOT encryption. For real security, restrict file system access to `mxLoreMCP.ini`.
## Architecture
- **Stack:** Delphi (RAD Studio) + TMS Sparkle + FireDAC + MariaDB
- **Protocol:** MCP 2024-11-05 (JSON-RPC 2.0 over HTTP+SSE)
- **Auth:** PBKDF2-HMAC-SHA256 API keys, per-project ACL
- **Units:** 43 Delphi units, dot-notation (`mx.*.pas`), 5 layers
## Directory Structure
```
mxLoreMCP.exe — Server (console)
mxLoreMCPGui.exe — Server (GUI with tray icon)
mxLoreMCP.ini — Configuration (create from .example)
mxLoreMCP.ini.example — Config template with documentation
sql/
setup.sql — Database schema (auto-run on first boot)
admin/
www/ — Admin UI (HTML/CSS/JS, served by the server)
lib/
libmariadb.dll — MariaDB client (or auto-detected from your MariaDB installation)
logs/ — Auto-created, log rotation at 5 MB
backups/ — Auto-created, mysqldump backups
```
## Skills (separate repo: [mxLore-skills](https://github.com/MicrotronX/mxLore-skills))
11 Claude Code skills for AI-assisted workflows:
| Skill | Purpose |
|-------|---------|
| `/mxOrchestrate` | Session orchestrator — workflows, ad-hoc tasks, team agents |
| `/mxSave` | Persist session state to DB |
| `/mxPlan` | Create implementation plans |
| `/mxSpec` | Write specifications |
| `/mxDecision` | Document architectural decisions (ADRs) |
| `/mxDesignChecker` | Review code/design against specs |
| `/mxBugChecker` | Find bugs with verification |
| `/mxHealth` | Knowledge DB consistency checks |
| `/mxSetup` | Developer onboarding (installs skills + proxy) |
| `/mxInitProject` | Bootstrap new project in DB |
| `/mxMigrateToDb` | Migrate local docs to MCP |
Install all via: `/mxSetup` in Claude Code.
## License
**Server:** BSL 1.1 (Business Source License) — see [LICENSE.txt](LICENSE.txt)
- Source available: read, review, modify
- Pre-built binaries: no Delphi or TMS license needed
- Non-commercial use in production: free
- Commercial use in production: requires a license from MicrotronX — contact info@microtronx.com
- After the change date (4 years), converts to Apache 2.0
**Skills:** MIT — fully open source, no restrictions. See [mxLore-skills](https://github.com/MicrotronX/mxLore-skills).
**Important:** To compile the server from source, you need RAD Studio (Delphi) and TMS Sparkle — both are commercial products. Most users should use the pre-built binaries from [GitHub Releases](https://github.com/MicrotronX/mxLore/releases), which have no such requirement.
### Third-Party
- `libmariadb.dll` — MariaDB Connector/C, LGPL 2.1. [Source](https://github.com/mariadb-corporation/mariadb-connector-c). Not bundled — auto-detected from your MariaDB installation.
## Why mxLore?
AI coding assistants forget everything between sessions. Your specs, decisions, lessons learned — gone. mxLore gives your AI a persistent knowledge base it can search, recall, and learn from. Self-hosted, your data stays yours.
**[See how it works at mxlore.dev](https://www.mxlore.dev)** — features, comparison, quick start guide.
## Links
- **[mxlore.dev](https://www.mxlore.dev)** — Features, demo, documentation
- [MicrotronX/mxLore](https://github.com/MicrotronX/mxLore) — Server source code
- [MicrotronX/mxLore-skills](https://github.com/MicrotronX/mxLore-skills) — Claude Code skills
- [MicrotronX](https://github.com/MicrotronX) — More tools by MicrotronX