Cortex MCP

Cortex MCP

Persistent brain, memory, loop controller, and reminder engine for AI coding agents.

Category
访问服务器

README

Cortex MCP

Persistent brain, memory, loop controller, and reminder engine for AI coding agents.

npm version License: MIT Node.js


What is Cortex?

Cortex is a local MCP server that gives AI coding agents persistent memory, task management, knowledge graphs, and a live dashboard. It runs on your machine, stores everything in SQLite, and never sends data to the cloud.

Agent starts → cortex_get_state → knows everything → works → cortex_save_snapshot → done

Features

31 MCP Tools

  • Core Loop — get_state, get_next_task, log_progress, check_reminders, save_snapshot
  • Project Setup — init (10-question onboarding), health check with auto-repair
  • Features & Files — track features, files, checkpoints, rollback
  • Tests & Issues — register tests, log bugs, resolve issues
  • Knowledge Base — dictionary, snippets, research, decisions, FTS5 search
  • Human Interaction — ask_human (pause for input), confirm_destructive (red modal)
  • V2 Advanced — token budget (180k), agent roles (4 levels), contradiction detection
  • Knowledge Graph — relationships between features, files, tests, issues

Live Dashboard

  • 11 tabs — Overview, Features, FileTree, Tests, Progress, Issues, Library, Research, Dictionary, Graph, Settings
  • D3 Knowledge Graph — force-directed visualization with 27+ nodes, drag/zoom/pan
  • WebSocket Live Push — dashboard updates instantly when data changes
  • Ctrl+K Search — global search across all tables
  • shadcn/ui — modern React 19 + Tailwind CSS interface

MCP Resources (10 endpoints)

Read project data without tool calls via cortex:// URIs: cortex://project, cortex://features, cortex://files, cortex://tests, cortex://issues, cortex://dictionary, cortex://progress, cortex://todos, cortex://relationships, cortex://snapshots

MCP Prompts (5 templates)

Pre-built session workflows: start-session, debug-issue, review-code, init-project, end-session

V2.1 Modules

  • Audit Trail — structured logging for every tool call
  • Episodic Memory — timestamped events with importance scoring
  • Context Compilation — full project state in 2ms

Quick Start

Option 1: npx (Recommended)

Add to your opencode.json:

{
  "mcp": {
    "cortex": {
      "type": "local",
      "command": ["npx", "-y", "@neuralnexustech/cortex-mcp@latest", "start", "--project", "."],
      "enabled": true,
      "env": {
        "CORTEX_PROJECT_PATH": "."
      }
    }
  }
}

Option 2: Local Install

npm install -g @neuralnexustech/cortex-mcp@latest

Then add to opencode.json:

{
  "mcp": {
    "cortex": {
      "type": "local",
      "command": ["cortex", "start", "--project", "/path/to/your/project"],
      "enabled": true,
      "env": {
        "CORTEX_PROJECT_PATH": "/path/to/your/project"
      }
    }
  }
}

Option 3: Clone & Run

git clone https://github.com/neuralnexustech/cortex-mcp.git
cd cortex-mcp
npm install
cd dashboard && npm install && npm run build && cd ..
node src/server.js

Session Lifecycle

START
  ↓
cortex_get_state        → project context (<200 tokens)
  ↓
cortex_get_next_task    → highest-priority pending todo
  ↓
WORK                    → write code, create files
  ↓
cortex_tick_file        → track each file created
  ↓
cortex_log_progress     → log completed work
  ↓
cortex_check_reminders  → handle warnings
  ↓
cortex_get_next_task    → next todo (or "ALL TASKS COMPLETE")
  ↓
...repeat...
  ↓
cortex_save_snapshot    → compress session to summary
  ↓
END

Tool Reference

Core Loop (5)

Tool Purpose
cortex_get_state Compressed project context
cortex_get_next_task Highest-priority pending todo
cortex_log_progress Log completed work
cortex_check_reminders 6 safety checks
cortex_save_snapshot Compress session summary

Project Setup (3)

Tool Purpose
cortex_init Initialize project (10-question onboarding)
cortex_set_active_project Switch between projects
cortex_health DB integrity + auto-repair

Features & Files (5)

Tool Purpose
cortex_add_feature Register a feature
cortex_update_feature Update feature status
cortex_tick_file Track file creation (auto-checkpoints)
cortex_get_files List all tracked files
cortex_rollback_file Restore file from checkpoint

Tests (2)

Tool Purpose
cortex_add_test Register a test
cortex_update_test Mark test passed/failed

Issues (2)

Tool Purpose
cortex_log_issue Log a bug or blocker
cortex_resolve_issue Mark issue resolved with fix

Knowledge Base (6)

Tool Purpose
cortex_write_dictionary Document a file/feature
cortex_get_detail Retrieve full dictionary entry
cortex_add_snippet Save reusable code snippet
cortex_add_research Log library research notes
cortex_add_decision Record architectural decision
cortex_search FTS5 + vector hybrid search

Human Interaction (2)

Tool Purpose
cortex_ask_human Pause, ask question, wait for answer
cortex_confirm_destructive Red confirmation modal for dangerous ops

V2 Advanced (5)

Tool Purpose
cortex_log_tokens Track token usage per action
cortex_get_token_budget Check remaining budget (180k default)
cortex_set_role Set agent role
cortex_get_role Get agent permissions
cortex_list_agents List all connected agents

Knowledge Graph (3)

Tool Purpose
cortex_add_relationship Link entities
cortex_check_contradictions Find conflicting data
cortex_resolve_contradiction Resolve conflicts

Skills (2)

Tool Purpose
cortex_get_skill Load SKILL.md guide
cortex_list_skills List available skills

Database

SQLite at .cortex/cortex.db with:

  • WAL mode — concurrent reads while writing
  • FTS5 — full-text search across all tables
  • 15+ tables — project, features, files, tests, issues, dictionary, progress, relationships, etc.
  • Triggers — auto-sync FTS index on insert/update/delete

Dashboard

Live at http://localhost:3001 when the server runs.

  • Ctrl+K — Global search overlay
  • Graph tab — D3 force-directed knowledge graph
  • LIVE indicator — WebSocket connection status
  • 11 tabs — Overview, Features, FileTree, Tests, Progress, Issues, Library, Research, Dictionary, Graph, Settings

REST API

Endpoint Description
GET /api/health Health check
GET /api/data All project data
GET /api/graph Knowledge graph (nodes + edges)
GET /api/search?q=<query> Search across tables
GET /api/audit Audit trail entries
GET /api/episodic Episodic memory events
GET /api/context Compiled project state
GET /ws-status WebSocket status
POST /human-answer Submit answer to pending question

MCP Resources

Read project data without tool calls:

URI Data
cortex://project Project config, features, todos
cortex://features All features with status
cortex://files All tracked files
cortex://tests All tests
cortex://issues All issues
cortex://dictionary File documentation
cortex://progress Recent activity
cortex://todos Pending tasks
cortex://relationships Knowledge graph edges
cortex://snapshots Session summaries

Architecture

cortex-mcp/
├── bin/cortex.js              # CLI entry point
├── src/
│   ├── server.js              # MCP server (main entry)
│   ├── api/server.js          # Express REST API + dashboard
│   ├── db/
│   │   ├── schema.js          # Table definitions + FTS5
│   │   ├── init.js            # DB connection (WAL mode)
│   │   ├── queries.js         # Read/write functions
│   │   └── cortex_v2.sql      # V2 migration
│   ├── tools/                 # 31 MCP tools
│   │   ├── state.js           # cortex_get_state
│   │   ├── tasks.js           # cortex_get_next_task
│   │   ├── search.js          # FTS5 + hybrid search
│   │   ├── contradictions.js  # Contradiction detection
│   │   ├── tokens.js          # Token budget
│   │   ├── roles.js           # Agent roles
│   │   ├── relationships.js   # Knowledge graph
│   │   ├── health.js          # Auto-repair
│   │   └── ...
│   ├── embeddings/            # ONNX vector engine
│   ├── websocket/server.js    # WebSocket live push
│   ├── audit/index.js         # Audit trail
│   ├── memory/
│   │   ├── episodic.js        # Episodic memory
│   │   └── compiler.js        # Context compiler
│   ├── resources/project.js   # MCP Resources
│   └── prompts/index.js       # MCP Prompts
├── dashboard/                 # React 19 + shadcn/ui
│   ├── src/
│   │   ├── App.jsx            # Main app + routing
│   │   ├── components/
│   │   │   ├── tabs/          # 11 tab components
│   │   │   ├── ui/            # shadcn/ui components
│   │   │   ├── Graph.jsx      # D3 knowledge graph
│   │   │   └── GlobalSearch.jsx
│   │   └── styles/globals.css # Tailwind + shadcn vars
│   └── dist/                  # Built dashboard
├── skills/                    # Agent skill files
├── docs/                      # Documentation
└── AGENTS.md                  # Agent guide

Environment Variables

Variable Default Description
CORTEX_PROJECT_PATH . Project root directory
CORTEX_API_PORT 3001 REST API port
CORTEX_PORT 4759 Dashboard port
CORTEX_SYNC_ENABLED false Enable cloud sync

Requirements

  • Node.js ≥ 18.0.0
  • Works on Windows, macOS, Linux

Links

Platform URL
Website https://neuralnexustech.com/
GitHub https://github.com/neuralnexustech/cortex-mcp
npm https://www.npmjs.com/package/@neuralnexustech/cortex-mcp

License

MIT © neuralnexustech

推荐服务器

Baidu Map

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。

官方
精选
JavaScript
Playwright MCP Server

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。

官方
精选
TypeScript
Magic Component Platform (MCP)

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。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

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。

官方
精选
本地
TypeScript
VeyraX

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。

官方
精选
本地
graphlit-mcp-server

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。

官方
精选
TypeScript
Kagi MCP Server

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。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

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。

官方
精选