{"id":"mcp-bridge","name":"MCP-Bridge","homepage":"https://github.com/SecretiveShell/MCP-Bridge","repo_url":"https://github.com/SecretiveShell/MCP-Bridge","category":"ai-ml","subcategories":["middleware","api-bridge","tool-integration"],"tags":["openai-api","middleware","bridge","fastapi","mcp-tools","llm-integration","sse","streaming","vllm","ollama"],"what_it_does":"Middleware that bridges the OpenAI API with MCP tools, allowing developers to expose MCP tool capabilities through an OpenAI-compatible API endpoint. Supports streaming and non-streaming chat completions with MCP tool calls, SSE bridge for external clients, REST API endpoints for MCP primitives, and optional API key authentication.","use_cases":["Expose MCP tools through an OpenAI-compatible API for existing OpenAI SDK users","Connect local inference engines (vLLM, Ollama) to MCP tool ecosystems","Build middleware that translates between OpenAI function calling and MCP tools","Provide SSE bridge for external MCP clients","Add API key authentication layer in front of MCP servers"],"not_for":["Direct MCP server implementation for specific services","Production-grade API gateway (no rate limiting, monitoring)","Non-OpenAI-compatible LLM APIs without adaptation"],"best_when":"You have existing OpenAI API workflows and want to add MCP tool support without rewriting your client code, or you need to bridge local LLM inference engines to MCP servers.","avoid_when":"You already have native MCP client support in your application, or you need a production-hardened API gateway with monitoring and rate limiting.","alternatives":["langchain-mcp-adapters","mcp-client-for-ollama"],"af_score":57.0,"security_score":55.0,"reliability_score":null,"package_type":"mcp_server","discovery_source":["github","github_awesome"],"priority":"low","status":"evaluated","version_evaluated":"0.5.1","last_evaluated":"2026-03-01T09:50:05.840969+00:00","performance":{"latency_p50_ms":null,"latency_p99_ms":null,"uptime_sla_percent":null,"rate_limits":null,"data_source":"llm_estimated","measured_on":null}}