GID MCP Server
Enables AI to analyze, query, and manage a graph-based representation of software architecture for impact analysis, dependency tracking, and design.
README
GID MCP Server
Model Context Protocol server for Graph-Indexed Development
Give AI assistants structural awareness of your codebase. GID represents software systems as typed, directed graphs — so AI can reason about architecture, not just syntax.

Why GID?
AI can generate code, but it can't answer:
- "What breaks if I change UserService?"
- "Which components implement the auth feature?"
- "What's the dependency path from Controller to Database?"
GID fills this gap by providing a graph-based map of your software architecture that AI assistants can query and update.
Two workflows:
- Top-down: Describe what you want to build → GID generates the architecture graph → AI implements against it
- Bottom-up: Extract a graph from existing code → Use it for impact analysis, safe refactoring, and planning new changes
The graph evolves with your project. Every time you add a feature or refactor, the graph updates — so AI always has the current map.
Dogfooding: GID's own architecture is defined as a GID graph. We used GID to build GID — tracking components, querying impact before refactoring, and planning new features. See the self-referential graph.
Tools
Query & Analysis
| Tool | Description |
|---|---|
gid_query_impact |
Analyze what components and features are affected by changing a node |
gid_query_deps |
Get dependencies or dependents of a node (with depth control) |
gid_query_common_cause |
Find shared dependencies between two nodes (useful for debugging) |
gid_query_path |
Find dependency path between two nodes |
gid_analyze |
Deep analysis of a file (functions, classes, complexity) |
gid_get_file_summary |
Structured file analysis for AI summarization |
gid_advise |
Graph health score, validation issues, and improvement suggestions |
gid_get_schema |
Get the GID graph schema with dynamic relations |
Graph Management
| Tool | Description |
|---|---|
gid_read |
Read graph structure (YAML, JSON, or summary) |
gid_init |
Initialize a new GID graph in a project |
gid_edit_graph |
Add, update, or delete nodes, edges, and relation types |
gid_refactor |
Rename, move, or delete nodes with cascade |
gid_history |
Version history — list, diff, or restore previous versions |
AI-Assisted
| Tool | Description |
|---|---|
gid_design |
Generate a graph from natural language requirements |
gid_extract |
Extract dependency graph from existing code (TypeScript/JavaScript) |
gid_semantify |
Propose semantic upgrades — map files to components, assign layers, detect features |
gid_complete |
Analyze docs to identify gaps and suggest graph additions |
gid_visual |
Generate interactive D3.js HTML visualization |
Resources
| Resource | Description |
|---|---|
gid://graph |
Current dependency graph (YAML) |
gid://health |
Health score and validation results |
gid://features |
List of all features in the graph |
Installation
Claude Desktop
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"gid": {
"command": "npx",
"args": ["graph-indexed-development-mcp"]
}
}
}
Claude Code
claude mcp add gid -- npx graph-indexed-development-mcp
Cursor / VS Code
Add to your MCP settings:
{
"gid": {
"command": "npx",
"args": ["graph-indexed-development-mcp"]
}
}
Quick Start
- Install the MCP server (see above)
- Initialize a graph in your project:
You: "Initialize a GID graph for this project"
→ Claude uses gid_init
- Extract dependencies from your code:
You: "Extract the dependency graph from the codebase"
→ Claude uses gid_extract
- Start querying:
You: "What would break if I change UserService?"
→ Claude uses gid_query_impact
You: "Design the architecture for a notification feature"
→ Claude uses gid_design
You: "Show me the project health score"
→ Claude uses gid_advise
Example Conversations
Top-Down: Design First, Then Build
You: "Design an e-commerce backend with auth, payments, and order tracking"
Claude uses gid_design →
Created 4 features: UserAuth, Payment, OrderTracking, ProductCatalog
Created 8 components across 4 layers
Created 15 dependency edges
Health score: 95/100
You: "Now implement the AuthService based on the graph"
Claude uses gid_query_deps →
AuthService depends on: UserRepository, TokenManager
Implements: UserAuth feature
Layer: application
Claude generates code that fits the architecture.
Bottom-Up: Extract from Existing Code
You: "Extract the dependency graph from this project"
Claude uses gid_extract →
Found 42 files, 156 dependencies
Grouped into 12 components across 4 layers
You: "I need to refactor UserService. What would break?"
Claude uses gid_query_impact →
Direct dependents: AuthController, ProfileController, OrderService
Affected features: UserRegistration, OrderPayment
5 components impacted, 2 features at risk
You: "Why do OrderService and PaymentService keep failing together?"
Claude uses gid_query_common_cause →
Shared dependency: DatabaseService
Both services depend on it — that's likely the root cause.
Continuous: Keep the Graph Updated
You: "I just added a NotificationService. Update the graph."
Claude uses gid_edit_graph →
Added node: NotificationService (Component, application layer)
Added edges: depends_on EmailClient, implements Notifications feature
You: "Check the project health"
Claude uses gid_advise →
Health: 87/100
Warning: NotificationService has no tests
Warning: EmailClient has 6 dependents (high coupling)
Suggestion: Consider splitting EmailClient into smaller modules
Visualization
You: "Visualize the current project architecture"
Claude uses gid_visual → Generates an interactive D3.js HTML file

Graph Format
GID uses a YAML-based graph format (.gid/graph.yml):
nodes:
UserAuth:
type: Feature
description: User authentication and authorization
priority: core
status: active
AuthService:
type: Component
layer: application
description: Handles authentication logic
path: src/services/auth.ts
AuthController:
type: Component
layer: interface
path: src/controllers/auth.ts
edges:
- from: AuthService
to: UserAuth
relation: implements
- from: AuthController
to: AuthService
relation: depends_on
Node types: Feature, Component, Interface, Data, File, Test, Decision
Relation types: implements, depends_on, calls, reads, writes, tested_by, defined_in, enables, blocks, requires, precedes, refines, validates, related_to, decided_by — plus custom relations you define.
Task Tracking
Nodes can have an optional tasks field for inline step tracking:
webhook-push:
type: Component
layer: infrastructure
status: in_progress
description: "Webhook push notifications HMAC-SHA256"
tasks:
- "[x] Implement HMAC signing"
- "[x] DM webhook events"
- "[ ] Run migration 011 on prod"
- "[ ] Add retry logic"
Convention: When all tasks are done, remove the tasks field and set status: active.
Tools
| Tool | Description |
|---|---|
gid_tasks |
Query tasks across the graph. No args = all pending. --node <id> for specific node. --done to include completed. |
gid_task_update |
Toggle task completion: --node <id> --task "task text" --done true/false |
gid_read |
Now shows tasks inline in summary output |
Display Format
webhook-push [Component, infrastructure, in_progress]
"Webhook push notifications HMAC-SHA256"
Tasks: 2/4 done
✅ Implement HMAC signing
✅ DM webhook events
☐ Run migration 011 on prod
☐ Add retry logic
Requirements
- Node.js >= 20.0.0
Related
- GID Methodology — Specification, examples, and dogfood graph
- GID CLI — Command line interface
- GID Paper — Formal methodology (Zenodo)
License
AGPL-3.0 — See LICENSE for details.
For commercial licensing, see COMMERCIAL-LICENSE.md.
Author
Toni Tang — @tonioyeme
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。