BogDB.Mcp.Codegen.Server 1.0.0

BogDb MCP Codegen Server

BogDb.Mcp.Codegen.Server is a specialized MCP server for code-intelligence agents (Codex, Claude, etc.).
It layers semantic, task-oriented tools on top of a purpose-built code-intelligence graph schema inside BogDB.

Why a separate server?

The base BogDb.Mcp.Server provides generic database tools (bogdb_query, bogdb_schema, etc.) that work against any BogDb database. A codegen agent needs more — it needs to ask questions like:

• "Where is UserService.CreateUser defined?"
• "What calls this function across all repos?"
• "If I change this protobuf schema, what services break?"
• "Who owns the payments service?"
• "What code is behind the enable_v2_payments feature flag?"

These require a pre-defined graph schema with nodes for repos, packages, modules, files, symbols, services, APIs, owners, etc., and semantic tools that generate the right Cypher internally.

Architecture

┌────────────────────────────────────────┐
│         Codegen Agent (Codex)          │
└─────────────┬──────────────────────────┘
              │  MCP (stdio)
┌─────────────▼──────────────────────────┐
│    BogDb.Mcp.Codegen.Server             │
│  ┌───────────────────────────────┐     │
│  │  11 codegen read tools        │     │
│  │  1 ingestion tool             │     │
│  └──────────┬────────────────────┘     │
│  ┌──────────▼────────────────────┐     │
│  │  4 base BogDb tools            │     │
│  │  (bogdb_query, schema, etc.)   │     │
│  └───────────────────────────────┘     │
└─────────────┬──────────────────────────┘
              │
┌─────────────▼──────────────────────────┐
│       BogDB (codegen schema)         │
└────────────────────────────────────────┘

Tool Catalog

Base BogDb Tools (pass-through)

Codegen Read Tools

Ingestion Tools

Graph Schema

Node Tables (15)

Repo, Package, Module, File, Symbol, Service, ApiEndpoint, DataSchema, Consumer, Owner, DeployUnit, FeatureFlag, Migration, Runbook, ArchDoc

Relationship Tables (16+)

CONTAINS_PACKAGE, CONTAINS_MODULE, CONTAINS_FILE, DEFINES_SYMBOL, REFERENCES_SYMBOL, DEPENDS_ON, EXPOSES_API, GOVERNED_BY_SCHEMA, CONSUMES_API, REPO_OWNED_BY, SERVICE_OWNED_BY, PACKAGE_OWNED_BY, DEPLOYED_AS, GATED_BY, HAS_MIGRATION, SERVICE_DOCUMENTED_IN, SERVICE_ARCH_DOC, SYMBOL_DOCUMENTED_IN, API_DOCUMENTED_IN, LINKED_TO

Run Locally

# In-memory database (default)
dotnet run --project BogDb.Mcp.Codegen.Server/BogDb.Mcp.Codegen.Server.csproj

# Persistent database
CODEGEN_DB_PATH=/path/to/codegen.bogdb dotnet run --project BogDb.Mcp.Codegen.Server/BogDb.Mcp.Codegen.Server.csproj

MCP Client Configuration

Claude / Codex (stdio)

{
  "mcpServers": {
    "bogdb-codegen": {
      "command": "dotnet",
      "args": ["run", "--project", "/path/to/BogDb.Mcp.Codegen.Server/BogDb.Mcp.Codegen.Server.csproj"],
      "env": {
        "CODEGEN_DB_PATH": "/path/to/codegen.bogdb"
      }
    }
  }
}

Typical Agent Workflow

1. Check graph status: codegen_schema_status → verify schema is initialized
2. Ingest repos: codegen_ingest_repo({repoPath: "/path/to/myrepo"}) → populate the graph
3. Query: Use semantic tools (codegen_find_symbol, codegen_callers, etc.) for code intelligence
4. Escape hatch: Use bogdb_query for ad-hoc Cypher when semantic tools aren't enough

Supported Languages (Ingestion)

Note: Day-1 extraction uses convention-based regex patterns. The interface is designed for future tree-sitter or LSP-based extraction upgrades.