"Automating Testing and Debugging for LLMs in Production" matching MCP servers:

A hexagonal TypeScript starter for production MCP servers with built-in authentication, billing, and tool scaffolding.

Enables AI agents to safely call enterprise tools through a governed MCP gateway with permission enforcement, blast-radius controls, input validation, and a full audit trail for every invocation.

Enables MCP clients to discover and execute operational tools such as database mutations and system diagnostics over stdio, with deterministic structured outputs and Langfuse observability.

A TypeScript MCP server starter kit supporting stdio and HTTP/SSE transports. Includes tools for system metrics, HTTP fetching with URL restrictions, and temporary key-value memory with namespace and TTL support.

Provides a production-oriented MCP server exposing tools, resources, and prompts over Streamable HTTP with JWT authentication and per-tool RBAC, enabling Amazon Bedrock Claude clients to securely invoke MCP capabilities.

Enables Codex to rank available tools and skills, rerank code and documentation search results, and triage saved output artifacts using Jev-based relevance scoring while leaving final decisions to Codex.

An MCP relay that adds a use_jev tool to every connected server, letting the Jev model autonomously select and execute tool calls while the calling model supplies any required text values. It enables agents to delegate multi-step tool use across all configured MCP servers through a single unified interface.

Enables Codex to manage a local Claude Code companion through MCP, implementing a dual-agent workflow where Codex handles reasoning and review while Claude Code performs engineering tasks.

A production-grade MCP server designed for multi-tenant, authenticated, and observable AI agent systems, enabling secure tool execution across heterogeneous data sources.

Enables Claude Code to control Firefox through tools for tabs, navigation, clicks, typing, scrolling, screenshots, page reading, forms, JavaScript, and console/network logs.

Enables users to describe an LLM fine-tuning workload once and receive ranked, task-aware GPU recommendations across a dozen cloud providers based on required VRAM/FLOPs, cost, and runtime. It surfaces the cheapest, fastest, and balanced options through MCP so coding assistants can plan training hardware without manual price comparison.

Enables building and evaluating LangGraph agents over a protocol-level MCP server to measure failure rates across four agent topologies.

Enables ZCode to control a real Chrome browser, including tabs, navigation, page content, clicks, form filling, keyboard input, scrolling, screenshots, and JavaScript evaluation through a local MCP bridge.

Enables AI agents to drive real Google Chrome with stealth and captcha handling, using Patchright to avoid bot detection.

Enables local opencode agents to control a live Chrome browser via MCP tools, including tab management, JavaScript execution, clicking, form filling, page reading, screenshots, and console log retrieval.

Enables OpenAI-compatible clients to access external tools, APIs, repositories, and data sources through MCP with allowlisting, approval, and argument validation, deployable as a FastAPI service or Python library.

Enables LLMs to discover and invoke generated API actions as MCP tools and read context through MCP resources for list and get-by-id operations. It provides HTTP-based MCP access for clients such as Cursor and Claude without extra glue code.

Enables ChatGPT to delegate writing, analysis, second opinions, and Japanese text polishing to Gemini through a single ask_gemini tool, without copying text to Gemini's web interface. Self-hosted on Cloudflare with your own Gemini API key and Google login.

Enables Codex to read a project's kickoff card, acceptance criteria, AGENTS rules, handoff package, progress, Git trail, and next step, and to append validated, audited, idempotent progress events backed by a matching project-record Git commit. Access is scoped to one workspace and project via short-lived, revocable, hash-stored credentials that cannot reach members, keys, deletion, publication, or deployment.

Exposes Coze workflows as MCP tools, allowing AI clients to discover and execute workflows from a specified Coze workspace. It provides a simple way to integrate external skills into platforms like OpenClaw without needing local environment configurations.

Enables MCP clients to read a project's Git-backed kickoff, acceptance criteria, handoff package, progress, and next step, and to append validated, audited progress events as project records. It provides short-lived scoped credentials and fail-closed tool allowlisting for safe, resumable AI-assisted development.

Enables AI agents to manage live Firefox tabs and native tab groups via MCP tools over stdio, with exact-match verification and safety constraints. It does not read page content or send browser data remotely.

Enables AI agents to control a Playwright-driven browser through natural-language instructions, including page interaction and live browser task execution.

Enables AI agents to securely reach enterprise tools and other agents through a governed gateway with pinned tool definitions, delegated authorization, content inspection, human approval, and signed audit trails.

Enables Claude Code to control the user's real Firefox browser through a local MCP server, with tools for tabs, navigation, page reading, clicking, typing, forms, screenshots, and file uploads under per-site consent.

Lets coding agents and MCP clients read, navigate, and interact with pages in the user's already signed-in Firefox through background tabs, without touching or stealing focus from the tab being viewed.

Enables LLMs to interact with Dify AI's chat completion API, including conversation context support and a restaurant recommendation tool.

Enables autonomous agents and MCP-compatible clients to enforce policy-approved actions using risk tiers, reversible checkpoints, and token approval bounds.

An authless Cloudflare Workers server that provides tools to scrape, analyze, and answer questions about websites using Cloudflare's Browser Rendering and AI capabilities.

AgentPay — x402 crypto data gateway on Stellar. 10 live pay-per-call tools: token prices, whale activity, gas tracker, DeFi TVL, Fear & Greed, Dune queries, token security. Agents pay USDC on Stellar. No API keys. Budget-aware sessions.

Allows AI to interact with Autodesk Revit via the MCP protocol, enabling retrieval of project data and automation of tasks like creating, modifying, and deleting elements.

Token-saving MCP proxy. Hides your MCP servers' tool lists behind a search tool and shrinks large tool results (GitHub, Playwright, logs, test output), keeping the full result available on request. One-command setup for Claude Code, Codex, Claude Desktop and Cursor.

A Model Context Protocol server implementation that enables AI assistants to interact with Slack workspaces, allowing them to browse channels, send messages, reply to threads, add reactions, and retrieve user information.

Merx - TRON Resource Exchange Merx is the first TRON resource exchange that aggregates energy and bandwidth providers into a single MCP server. GitHub: https://github.com/Hovsteder/merx-mcp Hosted SSE: https://merx.exchange/mcp/sse npm: merx-mcp Website: https://merx.exchange Stats * 52 tools, 30 prompts, 21 resources * 7+ energy providers (CatFee, ITRX, PowerSun, Fe

Enables MCP-compatible clients to manage AI agent teams, including agents, Kanban issues, projects, goals, approvals, budgets, and raw API access.

MCP Server for Google Cloud Healthcare API enables Agentic AI for a variety of FHIR-based digital health solutions, from smarter clinical workflows for Health Systems to Pre-Auth frameworks for Payers!

Badgr lets AI agents launch GPU workloads, deploy inference endpoints, run models and execute AI workflows through MCP. Route workloads to available compute, track execution and costs, and retrieve verified results without managing GPU infrastructure.

Enables AI assistants to query and analyze AI agent sessions from observability providers like Shepherd (AIOBS) and Langfuse, allowing users to debug agent runs, compare sessions, track performance, and analyze LLM usage patterns.

Hey @roocode community! I'm thrilled to share a project born from my work with Roocode and the vision of an AI-powered development team: the Anubis MCP Server! This system is heavily inspired by Roocode and designed to orchestrate an AI development workflow based on agile methodology. It simulates

An MCP server implementation that integrates AI assistants with Langfuse workspaces, allowing models to query LLM metrics by time range.