MCP Tools | Model Context Protocol Server Directory

Gives your coding agent the captured console, network, replay and screenshot evidence for a bug.

run any ai model. compose agents, stack knowledge, connect tools. one api, pay per run.

Run 150+ AI apps — image, video, audio, LLMs, 3D and more. Browse, execute, stream results.

Federated registry of 91k+ agent skills and MCP servers with one-command signed installs, plus a fleet control plane that keeps skill bundles identical across Claude Code, Codex, Hermes and OpenCode. 47 MCP tools. MPL-2.0, free tier, self-hostable.

CapSolver MCP Server exposes CAPTCHA solving as a Model Context Protocol tool, allowing AI agents and browser automation scripts to retrieve solutions for image, audio, and reCAPTCHA challenges through a standardized interface. By wrapping the CapSolver API in an MCP endpoint, developers can invoke solving functions just like any other tool, keeping authentication tokens and workflow logic isolated from the automation code. The server supports common challenge types, returns structured results with confidence scores, and handles retries and fallback strategies internally, reducing the need for custom error handling. Because it adheres to the MCP specification, it can be chained with other tools such as DOM extractors or form fillers within the same agent pipeline, enabling end‑to‑end automation of sites that employ CAPTCHAs without breaking the agent’s stateless design. Compared to direct API calls, the MCP wrapper simplifies integration, provides uniform logging, and lets teams swap solvers or add middleware without rewriting the core automation logic.

...

answerLoops is an MCP server that connects AI agents to a self-hosted support knowledge base, exposing a focused set of tools for retrieval and ticket handling. Agents can search the knowledge base, fetch specific FAQ entries, list or inspect existing tickets, create new ones on behalf of users, and generate suggested answers grounded in stored content. The self-hosted model means data stays within your own infrastructure, which is useful for teams handling internal documentation or customer data that cannot leave the network. Integration follows the standard Model Context Protocol: point your MCP-compatible client at the server, register the available tools, and the agent can call them directly in conversation flows. Compared to cloud-based support APIs, answerLoops trades ecosystem breadth for control and simplicity, making it a practical fit for small-to-mid support teams that already run their own services and want AI-assisted responses without third-party data exposure.

Pangolinfo Amazon Reviews MCP is a read-only Model Context Protocol server that surfaces structured Amazon review data through a standardized tool interface. It is designed for product research workflows where rating distributions, complaint patterns, media attachments, and ASIN-to-ASIN comparisons matter more than scraping raw HTML. Under the hood, it delegates to Pangolinfo's review aggregation pipeline rather than hitting Amazon directly, which keeps the integration simple and avoids the brittle proxy rotation most Amazon scrapers require. Developers typically wire it into research agents, competitive analysis copilots, or due-diligence assistants that need to summarize what real buyers are saying about a product, not just its star average. Because the protocol surface is narrow and schema-driven, it composes well with other MCP servers (catalog lookup, pricing, search) without leaking implementation details into prompts. Compared with generic web-scraping MCPs, it trades flexibility for cleaner, pre-normalized review signals and a smaller attack surface, since the server never writes back or mutates state.

For developers building location-intelligence tools or demographic analysis pipelines, mapping ZIP codes to counties is notoriously messy due to boundary overlaps. The ZIP-County Crosswalk MCP solves this by providing a direct interface to HUD’s official crosswalk datasets. Unlike simple lookup tables that might fail when a ZIP code spans multiple jurisdictions, this tool handles population-overlap filtering. This allows your agent or application to determine exactly which county holds the majority share of a specific ZIP code's population, ensuring higher data integrity for spatial queries. It integrates seamlessly via the Model Context Protocol, making it easy to plug into LLM-based workflows that require precise geographic context without manual data cleaning. Whether you are automating census-based reporting or refining logistics models, this tool provides a standardized, authoritative way to resolve spatial ambiguities.

For developers managing complex microservices, the biggest bottleneck in AI-driven development is the gap between code generation and reliable testing. Signadot’s MCP server bridges this by giving AI agents direct control over Kubernetes sandboxes. Instead of just writing code, an agent can now provision isolated, lightweight environments that fork real cluster traffic to test specific service changes. This moves beyond simple unit tests; it allows agents to validate logic against actual dependencies in a live-like state without the overhead of a full staging deployment. By integrating this into your workflow, you can automate the entire lifecycle of environment creation, traffic routing, and validation. It essentially transforms an LLM from a coding assistant into a sophisticated DevOps engineer capable of managing ephemeral infrastructure to ensure production-ready merges.

PostSyncer is a specialized MCP server designed to bridge the gap between LLM reasoning and cross-platform social media management. Instead of manually switching tabs to publish content, developers can integrate this tool into their AI workflows to automate the entire lifecycle of a post—from initial drafting and scheduling to final deployment across major networks like X, LinkedIn, TikTok, and Mastodon. Unlike standard API integrations that require custom boilerplate for every platform, PostSyncer provides a unified interface through the Model Context Protocol. This allows your agent to treat social media distribution as a single, structured capability. It is particularly useful for building autonomous marketing agents or content pipelines where the AI needs to verify post formatting, schedule timing, and monitor engagement metrics without human intervention. By offloading the multi-platform authentication and API complexities to a remote MCP server, you can focus on high-level orchestration rather than low-level integration debt.

The SendGrid MCP server bridges the gap between LLM reasoning and production-grade email infrastructure. Instead of manually writing boilerplate API calls to manage communication workflows, developers can grant their AI agents direct, structured access to SendGrid’s v3 API. This toolset provides 58 specialized functions, enabling agents to handle everything from triggering transactional emails and managing dynamic templates to pulling real-time delivery analytics. For developers building automated support bots, marketing automation agents, or intelligent observability dashboards, this eliminates the friction of context-switching between a chat interface and the SendGrid dashboard. Unlike basic wrappers, this implementation allows an LLM to actually 'reason' about email performance—for instance, identifying a drop in open rates and immediately querying the API to diagnose template issues. It is an essential integration for anyone moving from simple prompt-based email generation to autonomous, data-driven communication management.

For developers building agentic workflows, the biggest bottleneck is often the 'context gap'—the difficulty of feeding high-quality, structured documentation into an LLM without massive scraping overhead. This MCP tool bridges that gap by acting as a specialized discovery and conversion layer. Instead of writing custom scrapers for every library or framework, you can point your agent to this tool to automatically locate and parse `llms.txt` files. It handles the heavy lifting: finding the most specific index, resolving links, and delivering content in clean Markdown via a single remote endpoint. Unlike generic web search tools that return noisy HTML, this tool provides high-signal, LLM-ready documentation. It’s ideal for RAG pipelines or autonomous coding agents that need to instantly understand a new codebase or API documentation without manual intervention or complex preprocessing.

As the Model Context Protocol ecosystem expands, the biggest friction point for developers isn't building servers, but discovering the right ones for specific workflows. MCP Finder solves this by providing a unified, federated search layer across the official MCP Registry, Smithery, and npm. Instead of manually hunting through disparate repositories, you can query a single interface to find tools for database connectivity, API orchestration, or local filesystem management. What sets this apart is its developer-centric output: it doesn't just point you to a repo; it provides ready-to-use installation snippets tailored to your specific client. Whether you are configuring Claude Desktop or building a custom agentic workflow, this tool streamlines the integration phase by bridging the gap between discovery and deployment. It effectively turns a fragmented landscape into a searchable, actionable library of capabilities.

The Wayback Machine MCP tool provides a programmatic bridge to the Internet Archive, allowing LLMs to access historical snapshots of web content. Instead of hitting dead links or 404 errors during web browsing tasks, your agent can query the CDX index to find the closest archived version of a URL. It supports granular filtering, history collapsing, and includes a Common Crawl fallback to ensure data availability even when direct archives are incomplete. For developers building RAG pipelines or web-scraping agents, this eliminates the fragility of real-time web dependencies. It is a remote, keyless integration, meaning you can connect it to your existing MCP-compliant environment via a simple URL without managing complex API credentials. This is particularly useful for training models on historical data or debugging how web structures have evolved over time.

The GetBirthChart MCP server exposes the GetBirthChart API to any MCP-compatible client, letting you compute whole-natal charts, planetary positions, Big Three, Moon/Rising signs, aspects, and synastry reports on demand. It's useful for astrological web apps, research notebooks, or CLI tools that need structured astrological data without hosting ephemeris math yourself. After adding the server (npm install @getbirthchart/mcp or clone the GitHub repo) and configuring it in your MCP client, you call the exposed tools with birth data (date, time, location) and receive JSON results ready for rendering or analysis. Compared to building the calculations from scratch, this saves weeks of astronomy and astrological programming while staying current with the underlying service's updates.

For developers working with LLMs, the 'context window fatigue' is real. Every new session feels like starting from zero, forcing you to re-explain your tech stack, coding standards, or project architecture. The th-memory-mcp addresses this by providing a persistent, local-first memory layer via the Model Context Protocol. Unlike cloud-based memory solutions that raise privacy concerns, this tool utilizes a local SQLite database to store user preferences, past technical decisions, and specific workflow patterns. It functions as a long-term knowledge base that your AI agent can query to retrieve relevant context dynamically. Integrating it into your development harness—like OpenCode—allows the model to evolve alongside your project. Instead of manual prompting, the AI learns from your usage history to provide more personalized, accurate code suggestions and architectural advice, all while keeping your data strictly on your own machine.

Trooth Network introduces a specialized MCP server designed to bridge the gap between AI reasoning and real-world corporate accountability. For developers building agentic workflows or research-heavy applications, this tool provides a standardized way to query a live, read-only repository of company trust metrics. Instead of relying on outdated training data or hallucinations regarding corporate ethics, your LLM can pull verified, timestamped data on identity verification, security protocols, privacy policies, and AI governance. Because it operates as a no-auth MCP server, integration is frictionless: you simply plug it into your existing IDE or agent environment to enable real-time fact-checking. Unlike general search tools that return noisy web results, Trooth delivers structured, source-backed insights specifically curated for risk assessment and due diligence. It transforms your AI from a creative assistant into a tool capable of performing objective third-party audits during complex decision-making processes.

Managing Percona PostgreSQL clusters in Kubernetes often involves a high cognitive load, juggling connection pooling via PgBouncer, Point-in-Time Recovery (PITR), and disaster recovery workflows. The mcp-percona-pg tool bridges the gap between LLM-driven orchestration and complex database operations. Instead of manually executing kubectl commands or navigating complex operator manifests, you can use this MCP server to delegate database tuning, backup management, and pool configuration directly through your AI coding assistant. It provides a standardized interface to interact with your Percona instances, making it particularly useful for SREs and backend engineers who want to automate routine maintenance or troubleshoot connectivity issues using natural language. Unlike generic database tools, this is purpose-built for the Percona ecosystem on K8s, ensuring that operations like scaling or recovery follow best practices by default.

LegalAIMCP is a specialized Model Context Protocol server designed to bridge the gap between general-purpose LLMs and the highly fragmented legal tech ecosystem. Instead of forcing developers to manually curate lists of legal software, this tool provides a structured, queryable directory of AI tools and MCP servers tailored specifically for legal workflows. The core value lies in its scoring engine, which evaluates tool recommendations based on specific practice areas and firm scale, ensuring the context provided to your agent is relevant to the user's actual operational needs. For developers building legal-tech copilots or internal firm assistants, this integrates seamlessly as a tool-calling capability, allowing your model to perform real-time market research and software vetting without leaving the chat interface. It moves beyond simple keyword searches by providing a contextualized decision-making layer for legal technology procurement and implementation.

For developers building AI-driven bioinformatics agents, the Noodle Biomedical Literature Discovery MCP addresses the critical challenge of hallucination in scientific domains. Unlike generic search tools that return unverified text snippets, this protocol enables models to perform structured, source-linked queries across specialized biomedical databases. It supports two primary workflows: direct literature retrieval with verifiable citations and semantic graph traversal to map relationships between biological entities. This makes it ideal for building RAG pipelines where precision is non-negotiable, such as drug discovery workflows or automated literature reviews. By integrating this MCP, your agent moves from simple pattern matching to grounded reasoning, allowing it to navigate complex scientific ontologies and provide a traceable audit trail for every claim it makes. It effectively bridges the gap between LLM reasoning and high-fidelity biological knowledge bases.

FlashDesk introduces a critical bridge between LLM-driven coding and remote GUI environments. While SSH is the standard for headless terminal operations, it falls short when your workflow requires interacting with visual applications or complex desktop interfaces. FlashDesk solves this by implementing the Model Context Protocol to allow Claude Code to perform direct desktop manipulation—including screenshots, mouse movements, keyboard input, and file transfers—on a remote machine. Unlike standard remote desktop tools that require manual human intervention, this MCP server allows the AI to 'see' and act upon a remote workspace as if it were local. Security is handled via a per-connection approval flow on the target machine, ensuring the LLM cannot hijack a system without explicit authorization. For developers managing remote build servers, testing UI components, or automating legacy desktop software via natural language, FlashDesk transforms Claude from a code generator into a functional remote operator.

For developers building security-focused AI agents, bridging the gap between LLM reasoning and standardized vulnerability taxonomies is a persistent challenge. The fetch-cwe-list MCP server solves this by exposing the fetch-cwe-list library directly as a set of executable tools within the Model Context Protocol ecosystem. Instead of relying on an LLM's potentially outdated internal training data regarding Common Weakness Enumerations, your agent can now perform real-time, programmatic lookups of the CWE database. This allows for much higher precision during automated code reviews, threat modeling, or security auditing workflows. Integration is straightforward for any MCP-compliant client like Claude Desktop, turning a generic chat interface into a specialized security assistant capable of mapping discovered patterns to official industry standards. It effectively shifts the agent from 'guessing' vulnerability types to 'verifying' them against authoritative sources.

OpenQR is an MCP-native implementation designed to bridge the gap between LLM reasoning and physical-world connectivity. Unlike standard QR generators that only output static images, this server provides a robust suite of 17 tools to manage the entire lifecycle of a code. For developers, the real value lies in the support for dynamic QR codes; you can update the underlying destination URL without changing the printed pattern, and track scan analytics directly through your agentic workflow. It integrates seamlessly into any MCP-compliant environment, allowing your AI to not just 'see' a QR code, but to create, edit, and monitor them via a unified interface. Whether you are automating marketing workflows or building complex IoT provisioning systems, OpenQR offers a more programmable approach than traditional REST-only libraries by exposing management logic directly to the model's context.