title: VedaApex emoji: 🚀 colorFrom: blue colorTo: indigo sdk: docker app_port: 7860 pinned: false

VedaApex MCP Server

This backend now runs as a FastAPI application with an MCP layer added on top. Your original REST routes continue to work, and Claude can call the same backend through MCP tools.

Related MCP server: Metaso Search MCP Server

What changed

• Existing FastAPI routes remain available at the original REST endpoints.

• MCP endpoints are exposed at /mcp for HTTP transport and /sse for SSE transport.

• Search, health, and chat endpoints are exposed as MCP tools.

Installation

cd c:/Users/heyhi/Downloads/backend
.\.venv\Scripts\python.exe -m pip install -r requirements.txt

Configuration

Copy the sample environment file and adjust the values:

copy .env.example .env

Important variables:

APP_NAME=VedaApex Search Aggregation
APP_VERSION=1.0.0
APP_ENV=development
PORT=7860
HOST=0.0.0.0

PEXELS_API_KEY=your_pexels_key_here
NASA_API_KEY=DEMO_KEY
WIKIMEDIA_API_KEY=

MCP_HTTP_PATH=/mcp
MCP_SSE_PATH=/sse

Run the server

.\.venv\Scripts\python.exe main.py

The API will be available at:

• REST docs: http://localhost:7860/api/v1/docs

• MCP HTTP: http://localhost:7860/mcp

• MCP SSE: http://localhost:7860/sse

Test MCP locally

curl http://localhost:7860/api/v1/health
curl http://localhost:7860/mcp

You can also run the test suite:

.\.venv\Scripts\python.exe -m pytest -q test_mcp.py

MCP tools available

• health_check — returns service health and enabled providers

• unified_search — runs the main intelligent search workflow

• browser_search — runs the browser-style search helper

• chat — sends a chat request to the configured LLM provider

Claude Desktop configuration

Use this configuration in your Claude Desktop config file:

{
  "mcpServers": {
    "vedaapex": {
      "type": "streamable-http",
      "url": "http://localhost:7860/mcp"
    }
  }
}

If you prefer SSE instead of HTTP transport, use:

{
  "mcpServers": {
    "vedaapex": {
      "type": "sse",
      "url": "http://localhost:7860/sse"
    }
  }
}

Notes

• Your existing REST API behavior is preserved.

• Authentication headers are forwarded for tool calls when present.

• If an endpoint needs an API key, pass it through the request headers or environment variables used by the app.

Docker

docker build -t vedaapex-search . && docker run -p 7860:7860 vedaapex-search

Render

See DEPLOYMENT.md for cloud deployment guides.

❓ FAQ

Q: Can I add my own provider?
A: Yes! Create a provider class and add to providers/. The router will automatically include it.

Q: How are conflicts handled (space + scientific)?
A: Primary provider is chosen based on which category matches first.

Q: What if all providers fail?
A: Returns empty results with error details in logs.

Q: Can I disable a provider?
A: Yes, set ENABLE_PEXELS=false in .env

Built for intelligent search! 🧠✨

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference