Feature improve mcp

.env

@@ -0,0 +1,85 @@
+# =============================================================================
+# Graphify Configuration (.env)
+# =============================================================================
+# Place this file in the project root directory. Graphify loads it automatically
+# at startup. Settings use the GRAPHIFY__ prefix with double-underscore nesting.
+#
+# Configuration priority (highest wins):
+#   1. CLI arguments (--provider, --endpoint, etc.)
+#   2. dotnet user-secrets
+#   3. .env file              ← You are here
+#   4. Environment variables  (GRAPHIFY__*)
+#   5. appsettings.local.json (wizard-saved)
+#   6. appsettings.json       (built-in defaults)
+# =============================================================================
+
+# ── Global Settings ──────────────────────────────────────────────────────────
+
+# Active AI provider: azureopenai, openai, ollama, copilotsdk
+GRAPHIFY__Provider=openai
+
+# Project folder to analyze (default: current directory)
+GRAPHIFY__WorkingFolder=.
+
+# Output directory (default: graphify-out)
+GRAPHIFY__OutputFolder=graphify-out
+
+# Export formats, comma-separated (default: json,html,report)
+# Available: json, html, svg, neo4j, ladybug, obsidian, wiki, surrealdb, report
+GRAPHIFY__ExportFormats=surrealdb
+
+# ── OpenAI (Compatible) Provider ─────────────────────────────────────────────
+# Works with any OpenAI-compatible endpoint: OpenAI API, OpenCode Zen,
+# LocalAI, LiteLLM, vLLM, Groq, Together AI, etc.
+
+# Base endpoint URL (required)
+GRAPHIFY__OpenAi__Endpoint=https://opencode.ai/zen/v1
+
+# API key (required)
+GRAPHIFY__OpenAi__ApiKey=sk-k3mky4zwHRNIgfTbfqtVRJN8J5HNPin7NbJRA9X74giDjIGE8vSGSrrpiUDTl6Ns
+
+# Model identifier, e.g. gpt-4o, gpt-4o-mini, deepseek-v4-flash-free
+GRAPHIFY__OpenAi__ModelId=opencode/deepseek-v4-flash-free
+
+# ── Azure OpenAI Provider ────────────────────────────────────────────────────
+
+# Azure resource endpoint (required)
+GRAPHIFY__AzureOpenAI__Endpoint=https://my-resource.openai.azure.com/
+
+# API key (required)
+GRAPHIFY__AzureOpenAI__ApiKey=sk-your-key-here
+
+# Model deployment name (required)
+GRAPHIFY__AzureOpenAI__DeploymentName=gpt-4o
+
+# Optional model ID override (defaults to DeploymentName)
+GRAPHIFY__AzureOpenAI__ModelId=gpt-4o
+
+# ── Ollama Provider (Local) ──────────────────────────────────────────────────
+
+# Ollama server URL (default: http://localhost:11434)
+GRAPHIFY__Ollama__Endpoint=http://localhost:11434
+
+# Model to use (default: llama3.2)
+GRAPHIFY__Ollama__ModelId=llama3.2
+
+# ── GitHub Copilot SDK Provider ──────────────────────────────────────────────
+# Authentication via GitHub Copilot CLI session (no API key needed)
+
+# Model to use (default: gpt-4.1)
+GRAPHIFY__CopilotSdk__ModelId=gpt-4.1
+
+# ── SurrealDB (MCP Backend) ──────────────────────────────────────────────────
+# Configure when using `graphify serve` with a remote SurrealDB instance.
+# When Endpoint is set, remote mode is used; otherwise embedded mode.
+
+# Remote SurrealDB endpoint (e.g., http://localhost:8000)
+GRAPHIFY__SurrealDb__Endpoint=http://localhost:8000
+
+# Authentication credentials
+GRAPHIFY__SurrealDb__Username=root
+GRAPHIFY__SurrealDb__Password=root
+
+# Namespace and database (defaults: graphify / codebase)
+GRAPHIFY__SurrealDb__Namespace=archsentinal
+GRAPHIFY__SurrealDb__Database=codebase

.github/copilot-instructions.md

@@ -20,7 +20,7 @@ graphify-dotnet/
 │   ├── Graphify/                # Core library (graph pipeline, data structures, export)
 │   ├── Graphify.Cli/            # CLI tool (System.CommandLine)
 │   ├── Graphify.Sdk/            # GitHub Copilot SDK integration
-│   ├── Graphify.Mcp/            # MCP stdio server
+│   ├── Graphify.Cli/Mcp/        # MCP server built into CLI
 │   └── tests/
 │       ├── Graphify.Tests/              # Unit tests (xUnit)
 │       └── Graphify.Integration.Tests/  # Integration tests
@@ -48,7 +48,7 @@ graphify-dotnet/
 - `src/Graphify/` — Core library (graph pipeline, data structures, export)
 - `src/Graphify.Cli/` — CLI tool (dotnet tool, System.CommandLine)
 - `src/Graphify.Sdk/` — GitHub Copilot SDK integration
-- `src/Graphify.Mcp/` — MCP stdio server (ModelContextProtocol)
+- `src/Graphify.Cli/Mcp/` — MCP server built into the CLI (ModelContextProtocol)
 - `src/tests/Graphify.Tests/` — Unit tests (xUnit)
 - `src/tests/Graphify.Integration.Tests/` — Integration tests
 
@@ -68,7 +68,7 @@ graphify-dotnet/
 
 ### Library / Application Projects
 - Target `net10.0` only (single target — this is a .NET 10 project, not multi-target)
-- Project naming: `Graphify`, `Graphify.Cli`, `Graphify.Sdk`, `Graphify.Mcp`
+- Project naming: `Graphify`, `Graphify.Cli`, `Graphify.Sdk`
 - Enable deterministic CI builds: `<ContinuousIntegrationBuild Condition="'$(GITHUB_ACTIONS)' == 'true'">true</ContinuousIntegrationBuild>`
 - Include `<InternalsVisibleTo>` for the corresponding test project where needed
 - `Graphify.Cli` includes `<PackAsTool>true</PackAsTool>` for NuGet dotnet-tool publishing

.gitignore

@@ -22,3 +22,8 @@ obj/
 
 # Local config (may contain secrets)
 appsettings.local.json
+.tokensave/*
+.claude/PRPs/reports/surrealdb-phase1-schema-report.md
+nupkg/*
+.claude/*
+.tokensave/*

.gitmodules

@@ -0,0 +1,3 @@
+[submodule "lib/surrealdb.net"]
+	path = lib/surrealdb.net
+	url = https://github.com/surrealdb/surrealdb.net.git

ARCHITECTURE.md

@@ -82,8 +82,8 @@ graphify-dotnet/
 │   └── PipelineRunner.cs           # Pipeline orchestration
 ├── src/Graphify.Sdk/               # GitHub Copilot SDK integration
 │   └── CopilotExtractor.cs         # Copilot-specific extractors
-├── src/Graphify.Mcp/               # MCP stdio server
-│   └── McpServer.cs                # ModelContextProtocol server
+├── src/Graphify.Cli/               # CLI + MCP server
+│   ├── Mcp/GraphTools.cs           # ModelContextProtocol tool definitions
 └── src/tests/                      # Test projects
     ├── Graphify.Tests/             # Unit tests
     └── Graphify.Integration.Tests/ # Integration tests

CLAUDE.md

@@ -0,0 +1,117 @@
+# graphify-dotnet — ECC Project Guide
+
+**Project:** AI-powered knowledge graph builder for codebases (.NET 10)  
+**Architecture:** Multi-project solution with pipeline-based design  
+**Status:** Active development (AST extraction, semantic analysis, graph export)
+
+## Quick Commands
+
+```bash
+# Build
+dotnet build
+
+# Test (all)
+dotnet test
+
+# Test (specific)
+dotnet test --filter "FullyQualifiedName~Graphify.Tests"
+
+# Run CLI
+dotnet run --project src/Graphify.Cli -- --help
+
+# Package
+dotnet pack src/Graphify.Sdk -o ./nupkg
+
+# Clean build artifacts
+dotnet clean
+```
+
+## Project Structure
+
+### Core Library (`src/Graphify/`)
+- **Pipeline**: File detection → extraction → graph building → clustering → analysis → export
+- **Models**: Immutable records for nodes, edges, analysis results
+- **Export**: JSON, HTML (vis.js), SVG, Wiki (Markdown), Obsidian, Neo4j, Ladybug graph database
+- **Extraction**: Hybrid AST + semantic analysis (configurable AI provider)
+- **Cache**: SHA256-based semantic cache for reproducible runs
+
+### CLI (`src/Graphify.Cli/`)
+- Entry point for command-line usage
+- Configuration wizard (interactive setup for AI provider + analysis folder)
+- Pipeline orchestration
+
+### SDK (`src/Graphify.Sdk/`)
+- GitHub Copilot SDK integration
+- Copilot-specific extractors
+
+### MCP Server (built into CLI)
+- `graphify serve` — Model Context Protocol server (stdio-based)
+- Exposes graph operations to Claude and other MCP clients
+- Tools defined in `src/Graphify.Cli/Mcp/GraphTools.cs`
+
+### Tests
+- **Unit tests** (`Graphify.Tests`): Component-level testing
+- **Integration tests** (`Graphify.Integration.Tests`): Pipeline & export validation
+
+## Design Principles
+
+- **Composition over inheritance**: Interfaces + DI instead of deep hierarchies
+- **Immutable data**: Records for thread-safety
+- **.NET idioms**: `IOptions<T>`, `ILogger<T>`, `async/await`, `CancellationToken`
+- **Type safety**: Nullable reference types enabled; strong typing throughout
+- **Pipeline pattern**: Each stage is an `IPipelineStage<TIn, TOut>` implementation
+- **Decoupled stages**: Output from one stage feeds into the next via well-defined models
+
+## Code Health
+
+**Language Settings** (see `Directory.Build.props`):
+- C# latest (13)
+- Nullable reference types: enabled
+- Implicit usings: enabled
+- Code analysis: enforced
+- Warnings as errors: nullable violations only
+
+**Key Dependencies**:
+- **QuikGraph**: Graph data structures & algorithms (stable, 2.x)
+- **ModelContextProtocol**: MCP protocol for Claude integration (pre-1.0, monitor for CVEs)
+- **Spectre.Console**: CLI output formatting (pre-1.0, track releases)
+- **System.CommandLine**: CLI parsing (stable, 2.x)
+- **Microsoft.Extensions.*** packages: Follow .NET 10 preview releases (10.x)
+
+## AI Provider Setup
+
+The project supports three extraction modes:
+
+1. **Azure OpenAI**: Enterprise-grade semantic analysis (recommended for production)
+2. **Ollama**: Local/self-hosted LLM for privacy-first analysis
+3. **None (AST-only)**: Fast, zero-config extraction (no semantic relationships)
+
+Configure via the interactive wizard: `dotnet run --project src/Graphify.Cli -- config`
+
+## Code Review Guidance
+
+**Required for C# changes**: `/code-review` (ecc:csharp-reviewer)
+
+**Focus areas**:
+- Nullable reference type correctness (no #nullable disable)
+- Pipeline stage contracts (ensure `IPipelineStage<TIn, TOut>` invariants)
+- DI registration (Microsoft.Extensions.DependencyInjection patterns)
+- Async correctness (ConfigureAwait, CancellationToken propagation)
+- Export path safety (prevent directory traversal attacks)
+
+**Known constraints**:
+- MCP server uses stdio (not TCP) — test message framing end-to-end
+- Semantic cache keys are SHA256(file path + content) — invalidate on API changes
+- QuikGraph uses ref-based edges — be careful with value-type copies
+
+## Integration Points
+
+- **GitHub Actions**: CI build & publish workflows (see `.github/workflows/`)
+- **MCP Clients**: Claude, IDEs, chat applications via model context protocol
+- **Export formats**: Neo4j, Obsidian, vis.js visualization, custom wikis
+
+## Getting Help
+
+- See `docs/getting-started.md` for step-by-step walkthrough
+- See `ROADMAP.md` for planned features
+- See `ARCHITECTURE.md` for deep technical details

docs/.env.example

@@ -0,0 +1,85 @@
+# =============================================================================
+# Graphify Configuration (.env)
+# =============================================================================
+# Place this file in the project root directory. Graphify loads it automatically
+# at startup. Settings use the GRAPHIFY__ prefix with double-underscore nesting.
+#
+# Configuration priority (highest wins):
+#   1. CLI arguments (--provider, --endpoint, etc.)
+#   2. dotnet user-secrets
+#   3. .env file              ← You are here
+#   4. Environment variables  (GRAPHIFY__*)
+#   5. appsettings.local.json (wizard-saved)
+#   6. appsettings.json       (built-in defaults)
+# =============================================================================
+
+# ── Global Settings ──────────────────────────────────────────────────────────
+
+# Active AI provider: azureopenai, openai, ollama, copilotsdk
+GRAPHIFY__Provider=openai
+
+# Project folder to analyze (default: current directory)
+GRAPHIFY__WorkingFolder=.
+
+# Output directory (default: graphify-out)
+GRAPHIFY__OutputFolder=graphify-out
+
+# Export formats, comma-separated (default: json,html,report)
+# Available: json, html, svg, neo4j, ladybug, obsidian, wiki, surrealdb, report
+GRAPHIFY__ExportFormats=json,html,report
+
+# ── OpenAI (Compatible) Provider ─────────────────────────────────────────────
+# Works with any OpenAI-compatible endpoint: OpenAI API, OpenCode Zen,
+# LocalAI, LiteLLM, vLLM, Groq, Together AI, etc.
+
+# Base endpoint URL (required)
+GRAPHIFY__OpenAi__Endpoint=https://api.openai.com/v1
+
+# API key (required)
+GRAPHIFY__OpenAi__ApiKey=sk-your-key-here
+
+# Model identifier, e.g. gpt-4o, gpt-4o-mini, deepseek-v4-flash-free
+GRAPHIFY__OpenAi__ModelId=gpt-4o
+
+# ── Azure OpenAI Provider ────────────────────────────────────────────────────
+
+# Azure resource endpoint (required)
+GRAPHIFY__AzureOpenAI__Endpoint=https://my-resource.openai.azure.com/
+
+# API key (required)
+GRAPHIFY__AzureOpenAI__ApiKey=sk-your-key-here
+
+# Model deployment name (required)
+GRAPHIFY__AzureOpenAI__DeploymentName=gpt-4o
+
+# Optional model ID override (defaults to DeploymentName)
+GRAPHIFY__AzureOpenAI__ModelId=gpt-4o
+
+# ── Ollama Provider (Local) ──────────────────────────────────────────────────
+
+# Ollama server URL (default: http://localhost:11434)
+GRAPHIFY__Ollama__Endpoint=http://localhost:11434
+
+# Model to use (default: llama3.2)
+GRAPHIFY__Ollama__ModelId=llama3.2
+
+# ── GitHub Copilot SDK Provider ──────────────────────────────────────────────
+# Authentication via GitHub Copilot CLI session (no API key needed)
+
+# Model to use (default: gpt-4.1)
+GRAPHIFY__CopilotSdk__ModelId=gpt-4.1
+
+# ── SurrealDB (MCP Backend) ──────────────────────────────────────────────────
+# Configure when using `graphify serve` with a remote SurrealDB instance.
+# When Endpoint is set, remote mode is used; otherwise embedded mode.
+
+# Remote SurrealDB endpoint (e.g., http://localhost:8000)
+GRAPHIFY__SurrealDb__Endpoint=http://localhost:8000
+
+# Authentication credentials
+GRAPHIFY__SurrealDb__Username=root
+GRAPHIFY__SurrealDb__Password=your-password
+
+# Namespace and database (defaults: graphify / codebase)
+GRAPHIFY__SurrealDb__Namespace=graphify
+GRAPHIFY__SurrealDb__Database=codebase

docs/blog-post.md

@@ -60,11 +60,21 @@ dotnet run --project src/Graphify.Cli -- benchmark graphify-out/graph.json
 **Export to multiple formats:**
 
 ```bash
-dotnet run --project src/Graphify.Cli -- run . --format json,html,svg,neo4j,obsidian,wiki,report
+dotnet run --project src/Graphify.Cli -- run . --format json,html,svg,neo4j,obsidian,wiki,report,surrealdb
 ```
 
 The HTML export gives you an interactive vis.js graph. Click nodes, search by concept, filter by community. The Neo4j export lets you load the entire graph into a real graph database.
 
+**Query through AI assistants with the MCP server:**
+
+```bash
+# Build the graph first, then run the MCP server
+dotnet run --project src/Graphify.Cli -- run . --format json
+graphify serve graphify-out/graph.json
+```
+
+Connect Claude Desktop or VS Code and ask "Find the shortest path between the auth module and the database." The server exposes 5 tools for searching, navigating, and analyzing your codebase graph. See the [MCP Server docs](mcp-server.md) for client setup.
+
 ## Key Features
 
 - **Multi-language parsing**: Python, TypeScript, JavaScript, Go, Rust, Java, C#, C++, and more via tree-sitter AST extraction