MCP Protocol Integration

Vectora implements Model Context Protocol (MCP), a standardized protocol that allows any IDE with MCP support to connect and use Vectora as a context server. Works natively in Claude Code and Cursor.

What is MCP?

MCP is a standardized protocol for connecting AI models to external tools. Vectora exposes its functions (search, analysis) as MCP tools that any compatible IDE can use.

IDE (Claude Code / Cursor / other with MCP)
    ↓
  MCP Protocol (JSON-RPC)
    ↓
Vectora Server (localhost:9090 or remote)
    ↓
  12 Available Tools

Quick Start

Prerequisites

Node.js 18+
Vectora installed: npm install -g @kaffyn/vectora
API keys (Gemini, Voyage)
IDE with MCP support (Claude Code, Cursor, etc)

Step 1: Initialize Project

cd ~/your-project
vectora init --name "Your Project"

Step 2: Configure MCP in Your IDE

Config file (location varies by IDE):

Claude Code: ~/.claude/claude_desktop_config.json
Cursor: ~/.cursor/cursor_config.json
Other IDEs: See MCP documentation

Add Vectora:

{
  "mcpServers": {
    "vectora": {
      "command": "vectora",
      "args": ["mcp"],
      "env": {
        "GEMINI_API_KEY": "your-value",
        "VOYAGE_API_KEY": "your-value",
        "VECTORA_NAMESPACE": "your-namespace"
      }
    }
  }
}

Step 3: Test

Restart your IDE
Look for search_context tool in MCP menu
Test: @vectora search_context "How to validate tokens?"

12 Available Tools

Tool	Function
`search_context`	Semantic search for chunks
`search_tests`	Find related tests
`analyze_dependencies`	Find function callers
`find_similar_code`	Find similar code patterns
`get_file_structure`	Summarize file structure
`list_files`	List indexed files
`list_namespaces`	List namespaces
`get_namespace_stats`	Namespace statistics
`index_status`	Index status
`reindex`	Force re-indexing
`get_config`	Get current config
`get_metrics`	Execution metrics

Practical Workflows

Workflow 1: Understand Feature

You: "Explain how authentication works"
IDE: @vectora search_context "authentication"
Vectora: Returns relevant chunks
IDE: Shows chunks in context

Workflow 2: Debugging

You: "Why does this test fail?"
IDE: @vectora search_context "test X"
IDE: @vectora analyze_dependencies "tested function"
Vectora: Returns relevant context

Workflow 3: Code Review

You: "Review this function"
IDE: @vectora find_similar_code "your code"
Vectora: Finds similar patterns
IDE: Compares with existing code

Advanced Configuration

Custom Namespace

{
  "mcpServers": {
    "vectora": {
      "env": {
        "VECTORA_NAMESPACE": "staging" // Use different namespace
      }
    }
  }
}

Multiple Synchronized IDEs

If using multiple IDEs, both point to same config and namespace:

// Claude Code
{
  "mcpServers": {
    "vectora": {
      "command": "vectora",
      "args": ["mcp"],
      "env": {
        "VECTORA_NAMESPACE": "your-namespace"
      }
    }
  }
}

// Cursor - same config
{
  "mcpServers": {
    "vectora": {
      "command": "vectora",
      "args": ["mcp"],
      "env": {
        "VECTORA_NAMESPACE": "your-namespace"
      }
    }
  }
}

Both see the same chunks, indices, namespaces.

Troubleshooting

“Vectora command not found”

# Check installation
npm list -g @kaffyn/vectora

# Reinstall if needed
npm install -g @kaffyn/vectora --force

“Connection refused”

Vectora is not running as server:

# Start manually
vectora mcp

# Or via config with custom port
{
  "env": {
    "VECTORA_MCP_PORT": "9091"
  }
}

“API key not found”

Check environment variables:

echo $GEMINI_API_KEY
echo $VOYAGE_API_KEY

# If empty, configure in .env or config JSON

Performance

Expected latency: ~300-500ms (network + APIs)
Local search: ~100ms (no APIs)
Cache: Results cached in .vectora/
Concurrent: Supports multiple IDEs pointing to same server

Compatible IDEs

IDE	Support	Status
Claude Code	Native MCP	Tested
Cursor	Native MCP	Tested
VS Code	No native MCP	Use own extension
Zed	MCP supported	Not tested
Neovim	MCP via plugin	Not tested

For VS Code, use VS Code Extension.

Next Steps

VS Code? → VS Code Extension
ChatGPT? → ChatGPT Plugin
Gemini? → Gemini API

Part of Vectora ecosystem · Open Source (MIT)

Gemini API