# CodeFlowGraph MCP Server
A Model Context Protocol (MCP) server that provides semantic search and analysis capabilities for codebases using vector embeddings and call graph analysis.
## Setup
1. Install dependencies:
```bash
pip install -e .
```
2. Ensure you have the required dependencies:
- chromadb
- sentence-transformers
- mcp[cli]
## Run
Start the MCP server:
```bash
python -m code_flow.mcp_server
```
## Test
Run the test suite:
```bash
pytest tests/mcp_server/
```
Use the client for handshake and tool testing:
```bash
python client.py
```
## Config
Edit `code_flow/mcp_server/config/default.yaml` to customize:
- Watch directories for file changes
- Vector store settings
- Analysis parameters
- Drift detection settings
## Tools
The server provides the following tools:
### ping
Echo a message back.
```json
{
"input": {
"message": "hello"
},
"output": {
"status": "ok",
"echoed": "hello"
}
}
```
### semantic_search
Search functions semantically using natural language queries.
```json
{
"input": {
"query": "functions that handle user authentication",
"n_results": 5,
"filters": {}
},
"output": {
"results": [
{
"metadata": {
"name": "authenticate_user",
"file_path": "/path/to/auth.py"
}
}
]
}
}
```
### get_call_graph
Get call graph in JSON or Mermaid format.
```json
{
"input": {
"fqns": ["module.function"],
"format": "json"
},
"output": {
"graph": {
"functions": {
"module.function": {
"name": "function"
}
}
}
}
}
```
### get_function_metadata
Get detailed metadata for a specific function.
```json
{
"input": {
"fqn": "module.function"
},
"output": {
"name": "function",
"file_path": "/path/to/file.py",
"complexity": 5,
"parameters": ["arg1", "arg2"]
}
}
```
### query_entry_points
Get all identified entry points in the codebase.
```json
{
"input": {},
"output": {
"entry_points": [
{
"name": "main",
"file_path": "/path/to/main.py"
}
]
}
}
```
### generate_mermaid_graph
Generate Mermaid diagram for call graph visualization.
```json
{
"input": {
"fqns": ["module.function"],
"llm_optimized": true
},
"output": {
"graph": "graph TD\nA --> B"
}
}
```
### check_drift
Run drift detection against the current analyzed codebase.
```json
{
"input": {},
"output": {
"report": {
"summary": {
"structural_findings": 1,
"topological_findings": 0
}
}
}
}
```
Notes:
- Requires `drift_enabled: true` in config.
- Uses the current analysis state; run after initial indexing completes.
### impact_analysis
Compute impacted nodes from a change set using the call graph.
```json
{
"input": {
"changed_files": ["src/services/user_service.py"],
"depth": 2,
"direction": "both",
"include_paths": false
}
}
```
Notes:
- `changed_files` is optional; if omitted, the server uses files modified since the last analysis.
- `include_paths=true` returns call paths explaining the impact.
### reinforce_memory
Create or reinforce a memory entry in Cortex.
```json
{
"input": {
"content": "Use snake_case for database columns",
"memory_type": "TRIBAL",
"tags": ["conventions"],
"scope": "repo"
}
}
```
### query_memory
Query Cortex memory with decay-aware ranking.
```json
{
"input": {
"query": "database naming conventions",
"n_results": 5,
"filters": {"memory_type": "TRIBAL"}
}
}
```
### list_memory
List Cortex memory entries with filters and pagination.
```json
{
"input": {
"filters": {"memory_type": "EPISODIC"},
"limit": 10,
"offset": 0
}
}
```
### forget_memory
Delete a memory by id.
```json
{
"input": {
"knowledge_id": "<uuid>"
}
}
```
### update_context
Update session context with key-value pairs.
```json
{
"input": {
"key": "value"
},
"output": {
"version": 1
}
}
```
### get_context
Retrieve current session context.
```json
{
"input": {},
"output": {
"key": "value"
}
}
