The LORE MCP server provides architectural memory for AI coding assistants, enabling them to record, retrieve, and analyze technical decisions across sessions.
Core capabilities:
Record decisions (record_decision): Capture technical choices with description, reasoning, category (e.g., database, auth, security), who decided (AI or human), alternatives considered, and constraints that must not be violated
Retrieve context (get_context): Fetch all recorded decisions at session start to understand codebase rationale, with optional filtering by search query or category
Identify gaps (get_gaps): Surface decisions that were made but may not be fully implemented, highlighting inconsistencies or incomplete features
It integrates with AI coding tools like Claude Code and Cursor via MCP, preserving the "why" behind technical choices that AI assistants would otherwise lose between sessions.
Analyzes Git history to extract architectural insights, including bug-fix ratios, file churn, and commit quality.
LORE is a plugin-based code archaeology engine for TypeScript projects. It parses your AST, maps dependencies, detects circular deps, tracks async chains, scores type safety, and feeds architectural intelligence to AI coding assistants through the Model Context Protocol (MCP).
Built by EliotShift · Battle-tested on 16 real-world projects · 100% pass rate.
AI coding assistants write fast, but they lack architectural memory. They can't remember:
Which files are tightly coupled
Where circular dependencies live
Which types are spreading across boundaries
Which files change too often (hotspots)
What the architectural layers are
LORE solves this by giving your AI assistant a deep understanding of your codebase architecture — through CLI analysis and MCP server integration.
Related MCP server: Continuum
# | Analyzer | What It Does |
1 | AST Parser | Full TypeScript/TSX parsing with ts-morph + regex hybrid |
2 | Dependency Graph | Maps all imports, exports, re-exports across your project |
3 | Circular Dependency Detector | Finds cycles and ranks them by severity |
4 | Dependency Direction Checker | Enforces layer rules (e.g., no controller → DB imports) |
5 | Shannon Entropy | Complexity scoring per file (simple → very-complex) |
6 | Hotspot Analysis | Git-churn detection — files that change too often |
7 | Import Impact Analyzer | Shows the blast radius of every import |
8 | Type Safety Scorer | Grades your |
9 | Hidden Coupling Detector | Finds implicit dependencies through shared types |
10 | AI Recommendations | Prioritized fix suggestions (P0–P3) |
11 | Tooling Config Checker | Validates ESLint, Prettier, tsconfig settings |
12 | Breaking Change Detector | Flags high-risk deprecation patterns |
13 | Architectural Gap Finder | Identifies missing abstractions and patterns |
Expose LORE to Claude Desktop, VS Code, Cursor, or any MCP client:
Tool | Description |
| Full project analysis — scores, violations, complexity, hotspots |
| Health scores: overall, type safety, tooling, architecture |
| Circular deps, layer violations, architectural gaps |
| AI improvement suggestions (P0–P3) |
| Files with high git churn (red/yellow/green) |
| Shannon entropy complexity report |
| File imports, exports, consumers, complexity |
| Deep TS analysis: framework, layers, async chains, type flow |
Resource | Description |
| Latest analysis results (JSON) |
| Deep architecture graph (JSON) |
| Environment and cache status (JSON) |
lore [path] Analyze project (default: cwd)
lore analyze [path] Explicit analysis
lore init Extract architectural decisions
lore status View decisions by category
lore diff Diff against saved baseline
lore doctor Environment + tooling check
lore doctor --fix Auto-fix project setup
lore ignore List/manage ignore patterns
lore watch Watch + re-analyze on change
lore mcp inspect Inspect MCP server setup
lore mcp config Claude Desktop config snippet
lore version Show versionWebsite: eliotshift.github.io/lore-mcp
Page | Description |
Full feature overview, badges, and quick start | |
npm, npx, and Docker installation | |
All CLI commands: init, status, doctor, watch, diff, etc. | |
Claude Desktop, Cursor, and Windsurf setup | |
Real-world analysis of Express, NestJS, Next.js, and more | |
Common questions answered |
npm install -g lore-mcp# Analyze current project
lore
# Analyze a specific project
lore ./my-typescript-project
# Check environment and auto-fix
lore doctor --fix
# Watch for changes
lore watch --filter src/
# Diff from last baseline
lore diffAdd this to your Claude Desktop config:
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Linux: ~/.config/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"lore": {
"command": "npx",
"args": ["-y", "lore-mcp"]
}
}
}Or if installed globally:
{
"mcpServers": {
"lore": {
"command": "lore",
"args": ["mcp"]
}
}
}Restart Claude Desktop, then ask:
"Analyze my project architecture" → LORE runs
analyze_architecture"What are the circular dependencies?" → LORE runsget-violations"Which files are hotspots?" → LORE runsget-hotspots"What are the AI recommendations?" → LORE runsget-recommendations
lore-mcp/
├── src/
│ ├── algorithm/ # LoreGraph, AST parser, async chain builder, type tracker
│ ├── algorithms/ # AI recommendations, hotspot analysis, entropy, scoring
│ ├── analyzer/ # Dependency parsers, circular deps, direction, imports
│ ├── commands/ # CLI: doctor, diff, ignore, init, status, watch
│ ├── core/ # Plugin system, pipeline runner, graph engine
│ ├── lib/ # Hidden coupling, gaps, middleware chain, ontology
│ ├── mcp/ # MCP server + architecture bridge
│ ├── output/ # Formatter, markdown, SARIF, logger
│ ├── plugins/built-in/ # 13 built-in analysis plugins
│ ├── storage/ # Cache and decision store
│ ├── types/ # TypeScript type definitions
│ ├── cli.ts # CLI entry point
│ └── index.ts # MCP server entry point
├── package.json
└── tsconfig.jsonLORE uses a plugin-based architecture — every analyzer is a plugin:
interface LorePlugin {
name: string;
version: string;
analyze(context: AnalysisContext): Promise<PluginResult>;
}Built-in plugins include: circular-deps, coupling-matrix, dep-direction, entropy, gaps, hidden-coupling, hotspot, import-impact, middleware-chain, breaking-changes, type-safety, tooling-config, ai-recommendations.
Parse — LORE parses all .ts/.tsx files using a hybrid ts-morph + regex parser
Build Graph — Constructs a dependency graph with typed edges (import, type-ref, decorator, async-chain, implements, extends)
Run Plugins — 13 analyzers run in parallel through the plugin pipeline
Score — Computes health scores (type safety, tooling, architecture, overall 0–100)
Recommend — AI engine generates prioritized suggestions (P0 critical → P3 nice-to-have)
Serve — Results available via CLI output, MCP tools, or SARIF format
Project | Files | Result |
Express | 42 | 100% Pass |
Next.js | 68 | 100% Pass |
Fastify | 55 | 100% Pass |
NestJS | 38 | 100% Pass |
Prisma | 45 | 100% Pass |
Zod | 35 | 100% Pass |
TypeORM | 60 | 100% Pass |
All 16 test projects: 100% pass rate, zero crashes.
Node.js >= 18.0.0
TypeScript project (analyzes .ts and .tsx files)
Git (optional, for hotspot analysis)
ripgrep (optional, for faster file discovery)
Component | Technology |
Language | TypeScript 5.5+ |
AST Parsing | ts-morph 21 |
Protocol | Model Context Protocol (MCP) SDK 1.0 |
Transport | Stdio (Claude Desktop / IDE compatible) |
Validation | Zod schemas |
Output | ANSI terminal, Markdown, SARIF |
MIT © 2025 EliotShift
Get all architectural decisions for this project. Call this at the START of every session to understand the codebase.
| Name | Required | Description | Default |
|---|---|---|---|
| category | No | Filter by category (optional) | |
| query | No | Search for specific decisions (optional) |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden of behavioral disclosure. It mentions the tool retrieves decisions but doesn't describe key behaviors: whether it returns all decisions at once or paginates, what format the output is in, if there are rate limits, or authentication requirements. The instruction to call at session start implies it's foundational but lacks operational details.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is highly concise and front-loaded: two sentences with zero waste. The first sentence states the purpose, and the second provides usage guidance, both earning their place without redundancy or fluff.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given no annotations and no output schema, the description is incomplete for a tool that likely returns complex data (architectural decisions). It covers purpose and high-level usage but lacks details on output format, error handling, or behavioral constraints. For a retrieval tool with 2 parameters and no structured output info, this is minimally adequate but has clear gaps.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, so the schema already documents both optional parameters ('category' and 'query') with descriptions. The description doesn't add any meaning beyond what the schema provides—it doesn't explain how filtering or searching works, example categories, or query syntax. Baseline 3 is appropriate when the schema handles parameter documentation.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's purpose: 'Get all architectural decisions for this project.' It specifies the resource (architectural decisions) and scope (this project). However, it doesn't explicitly differentiate from sibling tools like 'get_gaps' or 'record_decision' beyond implying this is for retrieval rather than recording.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides explicit usage guidance: 'Call this at the START of every session to understand the codebase.' This gives clear context for when to use it (session initialization). However, it doesn't specify when NOT to use it or mention alternatives like the sibling tools, leaving some gaps in comparative guidance.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Find architectural gaps — decisions that were made but may not be fully implemented.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden. It mentions 'Find architectural gaps' but does not disclose behavioral traits such as what data sources are used, whether it's a read-only operation, potential side effects, or output format. This leaves significant gaps in understanding how the tool behaves.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single, efficient sentence that directly states the tool's purpose without unnecessary words. It is front-loaded and wastes no space, making it easy to parse quickly.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the complexity of finding architectural gaps, no annotations, and no output schema, the description is incomplete. It lacks details on how gaps are identified, what the output includes, or any behavioral context, making it insufficient for an agent to fully understand the tool's operation.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The tool has 0 parameters with 100% schema description coverage, so the schema fully documents the lack of inputs. The description does not add parameter details, but this is acceptable as there are no parameters to explain, maintaining clarity without redundancy.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's purpose with a specific verb ('Find') and resource ('architectural gaps'), explaining these as 'decisions that were made but may not be fully implemented.' It distinguishes from siblings like 'get_context' and 'record_decision' by focusing on gap identification rather than context retrieval or decision recording.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No explicit guidance is provided on when to use this tool versus alternatives. The description implies usage for finding architectural gaps but does not specify prerequisites, exclusions, or comparisons to sibling tools, leaving the agent to infer context.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Record an architectural decision. Call this whenever you make a technical choice that affects the codebase structure.
| Name | Required | Description | Default |
|---|---|---|---|
| category | Yes | Category: database, authentication, architecture, api, testing, deployment, security, performance, other | |
| decision | Yes | Short description of the decision made | |
| reason | Yes | Why this decision was made | |
| alternatives | No | Alternatives that were considered and rejected | |
| constraints | No | Constraints or rules that should not be violated | |
| author | No | Who made this decision |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries full burden. It states the action ('record') but doesn't disclose behavioral traits like how the decision is stored, if it's immutable, requires permissions, or has side effects. This is a significant gap for a tool with mutation implications.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single, efficient sentence that front-loads the purpose and usage without any wasted words. It's appropriately sized for the tool's complexity.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given no annotations and no output schema, the description adequately covers purpose and usage but lacks details on behavior, return values, or error handling. For a mutation tool with 6 parameters, it's minimally viable but has clear gaps in completeness.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, so the schema fully documents all parameters. The description adds no additional meaning beyond implying the tool is for recording decisions, which aligns with the schema but doesn't enhance parameter understanding. Baseline 3 is appropriate as the schema handles the heavy lifting.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the verb ('record') and resource ('architectural decision'), specifying it's for technical choices affecting codebase structure. It distinguishes from siblings like 'get_context' and 'get_gaps' by focusing on recording rather than retrieving, but doesn't explicitly differentiate beyond that.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
It provides clear context on when to use ('whenever you make a technical choice that affects the codebase structure'), which implicitly distinguishes it from sibling tools that likely retrieve information. However, it lacks explicit exclusions or named alternatives for similar recording tasks.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections. Dates show when Glama detected each change.
v0.1.5get_contextget_gapsrecord_decisionEach tool has a clearly distinct purpose: get_context retrieves existing decisions, get_gaps identifies implementation issues, and record_decision logs new choices. There is no overlap in functionality, making tool selection straightforward for an agent.
All tools follow a consistent verb_noun pattern (get_context, get_gaps, record_decision) with clear, descriptive names. The naming style is uniform and predictable throughout the set.
Three tools are appropriate for the server's purpose of managing architectural decisions, covering retrieval, gap analysis, and recording. While slightly minimal, each tool serves a distinct and necessary function without redundancy.
The tool set covers the core lifecycle of architectural decisions: retrieving context, identifying gaps, and recording new decisions. A minor gap exists in updating or deleting decisions, but agents can likely work around this for basic workflows.
Unclaimed servers have limited discoverability.
If you are the server author, to access and configure the admin panel.
Code intelligence platform for AI agents. 20 tools for architecture, security & impact analysis.
AI Agent with Architectural Memory. Impact analysis (free), tests and code from the graph (pro).
Shared memory for coding agents. Stop re-explaining your codebase every session.
Deterministic context layer for your codebase: change impact, blast radius, answers with receipts.
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/EliotShift/lore-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server
