MCP servers that ship as code — a repository or a package you install and run locally (usually over stdio). They have no hosted endpoint, so we cannot probe or grade them: instead each one shows its repository, install command, license, stars and when it was last updated.
Collected from GitHub repositories and npm, PyPI, Cargo, Go, Maven, NuGet and RubyGems packages tagged as MCP servers, the official MCP Registry, Docker's MCP catalog, ModelScope, Baidu MCP World, Cline and other public directories; one entry per repository. A server that also offers a hosted endpoint is listed under Remote instead. These entries are not counted in our reachability, verification or readiness statistics.
-
Drop-in OpenAI / Anthropic / Gemini BaseURL Proxy & Edge MCP Server (~0.04ms). Cut LLM input tokens by 60–99% via AST code skeletons, JSON pruning & stream healing with 1 line of config.
MIT ★ 1 -
proxyclawdgithub.com/dyshay/proxyclawdupdated this year
MITM proxy for Claude Code with real-time TUI, web UI & MCP server — intercept, inspect, and send messages to the Anthropic API. Headless mode for automation.
★ 15 -
Smart Context Engineering Gateway for AI agents: lossless tool-output compression, episodic memory folding and a reversible diet_ref vault (MCP server + OpenAI/Anthropic proxy + local dashboard)
MIT -
See where your AI agent's tokens go and stop paying for the ones that do nothing. Local CLI, MCP server, hook and proxy for Claude Code, Claude Desktop, Cursor, Codex, Gemini CLI and Anthropic/OpenAI/Gemini API apps. No API key, no telemetry.
npm MIT ★ 3 -
Local MCP server that gives coding agents a verifiable project memory, trims what enters their context, and masks secrets — with honest accounting of what it saves and costs.
npm MIT -
Keeps LLM prompt-cache prefixes stable so provider prompt caching actually hits: local proxy (point your client's baseURL at localhost) + MCP server. Fully local, no telemetry.
npm Apache-2.0 -
Cut Claude Code costs: Claude stays the brain while its subagents run on cheaper models (OpenRouter, DeepSeek, Qwen, GLM, Kimi). Rust CLI, local proxy and MCP server with secret/PII redaction, budgets and savings reports.
MIT ★ 1 -
Lightweight, single-binary, production-grade Rust LLM routing gateway & dashboard with health-aware failover, Claude Code compatibility, and MCP server.
Apache-2.0 ★ 4 -
llmtest-mcpgithub.com/tjacquesson/llmtest-mcpupdated this year
MCP server for LLMTest — benchmark AI models on real prompts, find cheaper alternatives across 340+ models. Works with Claude Code, Cursor, Windsurf.
MIT -
Hot-reload for MCP servers: edit, reload(), call again in the same session. Plus mount any of 130k+ community servers mid-session, Docker-caged.
npm MIT ★ 3 -
MCP-style HTTP server that multiplexes N database/REST backends behind one auth-protected endpoint - LRU cache, SQL allow-list, rate limiting, and per-target health checks.
MIT ★ 1 -
agentsgategithub.com/agentsgate/agentsgateupdated this year
Undo for AI agents: an MCP proxy that checkpoints every tool call before it runs, so a bad `DROP TABLE` or `rm -rf` rolls back in seconds. Risk scoring, human-in-the-loop approval, tamper-evident audit log, dashboard, and MCP servers for PostgreSQL/MySQL/SQLite. Local-first, MIT, TypeScript, Node 20+.
MIT -
Secure SQL proxy for AI agents — AST safety, per-agent RLS, audit log, MCP-native.
MIT ★ 1 -
Cryptographic proxy and credential vending for LLM applications. Zero secrets on disk. mTLS everywhere. Apache 2.0.
Apache-2.0 ★ 1 -
Docker container for centralized Ruflo MCP server with PostgreSQL (RuVector). Architecture: Ruflo (stdio) → Express proxy (Streamable HTTP /mcp).
★ 4 -
HTTP↔WebSocket bridge daemon between thunderbird-cli (or any HTTP client) and the Thunderbird WebExtension. Stateless proxy, localhost-only.
npm MIT -
@actualjob/mcp-servergithub.com/capitalthought/actualjobupdated this year
MCP server for actual-job.com — registers all jd_* tools and proxies to the Tool Gateway. Stdio transport for local agents (Claude Desktop, Cursor, Windsurf).
npm MIT -
Proxy MCP: one server that runs Playwright, Figma, Chrome DevTools, Shopify Dev and Context7 behind a single connection - and cuts their token cost with compact snapshots, caching and session recovery.
npm MIT ★ 24 -
sentinel-aigithub.com/MaxwellCalkin/sentinel-aiupdated this year
Real-time AI safety guardrails for LLM apps. 10 scanners: prompt injection, PII, harmful content, code vulnerabilities, obfuscation detection. Sub-ms latency. Python + TypeScript SDKs. MCP proxy. Claude Code hooks.
Apache-2.0 ★ 27 -
Self-hosted AI workspace that compresses prompts before they leave your machine — 79% fewer tokens, measured. Drop-in gateway for Claude Code, Cursor, Aider and any OpenAI/Anthropic client, plus an MCP server. Free, MIT, zero telemetry.
MIT ★ 2 -
@markcolabs/mcpgithub.com/mark57-ux/digitalcalculatorupdated this year
MCP stdio↔HTTPS shim for digitalcalculator.info financial calculators. Proxies MCP JSON-RPC to the hosted server at mcp.digitalcalculator.info/mcp. Per ADR-0044a: engine source is private; this package is a thin transport adapter only.
npm MIT -
Official IPGeolocation.io MCP server: 17 tools for accurate IP geolocation, company information, ASN and network data, IP security (VPN, proxy, Tor, bot, hosting and spam detection), abuse contacts, time zones, astronomy, currency data and user-agent parsing.
npm MIT ★ 6 -
Self-hosted AI API key manager and LLM gateway. Store OpenAI, Gemini, Claude, OpenRouter and Groq keys once, call any model by name. Key rotation, OpenAI-compatible proxy, web dashboard, and an MCP server with sandboxed Code Mode for Claude Code, Cursor and Codex.
MIT