Corn Intelligence Server

A high-performance Model Context Protocol (MCP) server integrated with a Django-powered analytics dashboard. Features the Corn Intelligence toolset for comprehensive AI interaction tracking, AST code analysis, and real-time token efficiency metrics.

🚀 Key Features

1. Intelligence Dashboard

• Real-time Monitoring: Every tool call is logged with latency, token usage, and efficiency metrics.
• Auto-Reload: Dashboard refreshes every 15 seconds with a visual countdown indicator.
• Token Analytics: Visualize direct token costs vs. context savings with efficiency scoring.
• Session Explorer: Browse and drill into past sessions at /sessions/.

2. Corn Intelligence Toolset (v2.0) — 19 Tools

#	Tool	Category	Description
1	corn_health	Core	System health — CPU, RAM, uptime
2	corn_session_start	Core	Begin a tracked work session
3	corn_session_end	Core	End session with summary
4	corn_memory_store	Memory	Store agent memory with tags
5	corn_memory_search	Memory	Keyword search over memories
6	corn_knowledge_store	Knowledge	Store a shared knowledge item
7	corn_knowledge_search	Knowledge	Semantic search over knowledge base
8	corn_code_read	Code	Read raw source code from any file path
9	corn_detect_changes	Code	Uncommitted git changes cross-referenced with AST graph
10	corn_list_repos	Code	List indexed repositories with symbol counts (auto-indexes on first call)
11	corn_code_search	Code	Hybrid AST symbol search by name
12	corn_code_context	Code	360° symbol view: callers, callees, hierarchy
13	corn_code_impact	Code	Blast radius analysis — which files depend on a given file
14	corn_cypher	Code	Graph-style queries: (a)-[:CALLS]->(b)
15	corn_tool_stats	Analytics	Usage analytics over last 50 tool calls
16	corn_quality_report	Quality	Submit a 3-dimension quality report (Clarity/Efficiency/Security)
17	corn_record_conversation	Analytics	Log raw conversation token usage to Dashboard
18	corn_plan_quality	Quality	Score a plan text against 8 quality criteria (must ≥80%)
19	corn_changes	Analytics	Check recent git commits by agents (git log -n 5)

3. AST Code Intelligence Engine

• Auto-Indexing: corn_list_repos automatically scans and indexes Python source files on first call.
• Symbol Graph: Stores Function, Class, and Method nodes with CALLS / INHERITS relationships in MySQL.
• Blast Radius: corn_code_impact identifies all callers transitively affected by changes to a file.

4. Token Efficiency Metrics

• Accurate Counting: Heuristic blends character density and word count — max(chars/4, words × 1.35) — for reliable estimation across code and natural language.
• Context Saved: For search tools (memory_search, knowledge_search, code_search), tokens saved = total DB tokens − tokens returned. Reflects real context reduction.
• Efficiency (%): Saved / (Used + Saved). Aim for >80%!

🏛️ Architecture

IDE (Antigravity / Claude)
    │  stdio (MCP protocol)
    ▼
Docker: personal-mcp-web
    ├── Django MCP Server (run_mcp management command)
    │       └── 19 Corn Intelligence Tools
    ├── AST Indexer (mcp_server/utils/indexer.py)
    └── Django Dashboard (http://localhost:8000)
            └── Real-time Activity Log, Sessions, Token Analytics
    │
    ▼
Docker: mcp_mysql_db (MySQL 8)
    ├── Session, ToolLog
    ├── Memory, Knowledge
    └── Repository, Symbol, SymbolRelation, PlanQuality

🏃 Getting Started

1. Launch Services:

   docker-compose up -d
   

2. Connect IDE — add to your MCP config (mcp_config.json):

   {
     "mcpServers": {
       "personal-mcp-v2": {
         "command": "/usr/local/bin/docker",
         "args": ["exec", "-i", "personal-mcp-web", "python", "manage.py", "run_mcp", "--verbosity", "0", "--no-color", "--skip-checks"]
       }
     }
   }
   

3. Run Migrations (first-time setup):

   docker exec personal-mcp-web python manage.py migrate
   

4. Explore Dashboard: http://localhost:8000/

🛠️ Development

Apply new migrations after model changes:

docker exec personal-mcp-web python manage.py makemigrations mcp_server
docker exec personal-mcp-web python manage.py migrate mcp_server

Check Django config inside container:

docker exec personal-mcp-web python manage.py check

View live logs:

docker logs -f personal-mcp-web

---

Powered by Antigravity — Premium AI Engineering.