DEV Community: CuratedMCP

MCP Ecosystem Week 20: Database Tooling Enters the Ring as Giants Dominate Discovery

curatedmcp — Mon, 11 May 2026 10:17:57 +0000

MCP Ecosystem Week 20: Database Tooling Enters the Ring as Giants Dominate Discovery

The MCP ecosystem continues its steady expansion with the addition of Local-YDB, an unofficial server that brings local YDB database management into the MCP fold. Meanwhile, the weekly trending list reveals a clear pattern: integration servers backed by major platforms—OpenAI, GitHub, Figma, and Anthropic—continue to capture developer attention at scale.

This Week in MCP

We added one new server this week:

Local-YDB unofficial mcp server (FREE): A community-built server for managing local YDB database instances. YDB, Yandex's distributed SQL database, joins the growing roster of database-specific MCP tools as developers seek to integrate data layer operations directly into their AI workflows.

The addition pushes the verified catalog to 70 servers, with 69 remaining free. No new playbooks were added this week, keeping the focus on server infrastructure expansion rather than orchestration patterns.

Trending Servers

GitHub Copilot MCP continues its dominant run, commanding 98K views this week. The server channels GitHub's code intelligence—completions, explanations, and review capabilities—into any MCP client, making it an obvious draw for developers working in Cursor and other AI IDEs. It's the bridge between GitHub's ecosystem and broader AI agent workflows.

OpenAI MCP lands at 87K views, offering official access to GPT-4o, DALL-E, Whisper, and Embeddings APIs through MCP. The steady demand signals that developers want to orchestrate multiple OpenAI capabilities within unified agent contexts rather than making isolated API calls.

Figma MCP rounds out the top three with 82K views, serving design engineers and teams building AI-assisted workflows around design systems. Direct access to design files, components, and design tokens from within an agent drastically reduces context-switching overhead—a compelling value proposition that the numbers clearly reflect.

The pattern here is worth noting: first-party integrations from well-known platforms capture the lion's share of attention. GitHub, OpenAI, Figma, and Anthropic collectively pull roughly 420K views across the top five slots. Developer friction and discoverability favor the names they already trust.

Quick Take

The 100% free-tier catalog (69 of 70 servers) masks an important reality: pricing structures around MCP are still forming. No commercial/premium servers have gained traction yet, suggesting the ecosystem remains in the "free tooling to build the foundation" phase. What's notable is how developers are using free servers—they're threading together specialized capabilities (database tools, design platforms, code hosts) into composite agents rather than adopting monolithic platforms. This suggests MCP's real value isn't in individual integrations, but in the ability to mix and match. The Local-YDB addition is a quiet marker: database teams, not just SaaS platforms, are beginning to see MCP as infrastructure worth supporting. Watch for more domain-specific tooling (observability, data, infrastructure) to follow as the pattern becomes clearer.

Browse all 70 verified servers at CuratedMCP.

PagerDuty MCP: Let AI Handle Your On-Call Chaos

curatedmcp — Sun, 10 May 2026 10:45:34 +0000

PagerDuty MCP: Let AI Handle Your On-Call Chaos

Operations teams spend too much time context-switching between incident dashboards, chat, and email. PagerDuty MCP is the official integration that gives Claude, Cursor, and other AI agents direct access to your incident response workflows—so they can create incidents, manage alerts, update on-call schedules, and keep responders aligned without human hand-offs.

What It Does

PagerDuty MCP bridges the gap between your AI agent and your incident management system. Instead of manually toggling between tools, you can now ask Claude to create an incident, check who's on-call, escalate a critical alert, or annotate a timeline—and the agent executes those commands directly in PagerDuty.

The server unlocks core capabilities: create and update incidents with custom urgency and priority levels, list active on-call users and rotation schedules, manage escalation policies and team assignments, query historical alerts and response timelines, annotate incidents with context or responder notes, and trigger incident actions programmatically from plain-language AI commands.

This is especially powerful for automation. When your monitoring system fires a critical alert, an AI agent can automatically open a PagerDuty incident, notify the right on-call engineer, and add diagnostic context—all in seconds. Or during an active incident, your agent can pull recent alert history, suggest escalation paths, and track response metadata without anyone leaving their terminal.

How to Install

npx -y @pagerduty/mcp

Add the server to your Claude Desktop config:

{
  "mcpServers": {
    "pagerduty-mcp": {
      "command": "npx -y @pagerduty/mcp",
      "env": {
        "PAGERDUTY_API_TOKEN": "your-api-token-here"
      }
    }
  }
}

You'll need a valid PagerDuty API token with appropriate scopes. Get yours from your PagerDuty account settings.

Real-World Use Cases

Incident triage at scale: Ask Claude to review your last 50 incidents, identify patterns (e.g., recurring database timeouts), and propose new escalation policies without manually digging through the UI.
Automated alert response: When a monitoring tool detects anomalies, an AI agent creates a PagerDuty incident, pulls the on-call roster, checks recent similar incidents, and sends a pre-populated summary to the responder.
On-call coordination: Query who's currently on-call across multiple services, check schedule conflicts, and let Claude suggest load-balancing changes based on response time trends.

Full install guides for Claude Desktop, Cursor, Windsurf, and more at CuratedMCP.

Splunk MCP: Query your observability stack directly from Claude

curatedmcp — Sat, 09 May 2026 10:45:37 +0000

Splunk MCP: Query your observability stack directly from Claude

Splunk MCP is the official Model Context Protocol server that bridges AI agents to your Splunk instance. Instead of context-switching between Claude and your monitoring platform, you can now ask your AI assistant to search logs, inspect alerts, correlate incidents, and surface root causes—all without leaving your chat.

What It Does

Splunk MCP gives Claude and other AI agents direct access to your observability and security data. You can run SPL (Splunk Processing Language) queries, retrieve event results, inspect logs across services, and access alert timelines. The server also lets you search saved dashboards, manage data inputs, and correlate security events with operational metrics.

The key unlock: your AI agent can now triage incidents autonomously. Feed it vague symptoms ("why did latency spike at 3pm?"), and it can query your Splunk instance, pull relevant logs and metrics, cross-reference alert history, and surface likely root causes. For SRE and SecOps teams drowning in data, this cuts investigation time significantly.

How to Install

Install via npm:

npx -y @splunk/mcp

Then add Splunk MCP to your Claude Desktop config:

{
  "mcpServers": {
    "splunk-mcp": {
      "command": "npx -y @splunk/mcp",
      "env": {
        "SPLUNK_URL": "https://your-splunk-instance.com",
        "SPLUNK_USERNAME": "your_username",
        "SPLUNK_PASSWORD": "your_api_token"
      }
    }
  }
}

Replace placeholders with your Splunk credentials and instance URL. Full setup guides for Cursor, Windsurf, and other clients available in the official docs.

Real-World Use Cases

Incident triage at speed: "Hey Claude, I got a PagerDuty alert for high error rates. Search Splunk logs from the last 30 minutes and tell me which service is failing." Claude queries your instance, correlates events, and suggests remediation steps.
Root cause analysis: "Why did our database queries slow down yesterday around 2pm?" Claude pulls metrics, compares baselines, and surfaces anomalies you'd normally miss.
Security event correlation: "Find all failed login attempts from the last 48 hours and correlate them with successful logins from unusual IPs." Claude runs complex multi-field searches without you writing SPL.

Full install guides for Claude Desktop, Cursor, Windsurf, and more at CuratedMCP.

ServiceNow MCP: Automate ITSM workflows without leaving your AI agent

curatedmcp — Fri, 08 May 2026 10:45:35 +0000

ServiceNow MCP: Automate ITSM workflows without leaving your AI agent

ServiceNow MCP is the official integration that lets Claude, Cursor, and other AI agents interact directly with ServiceNow's ITSM platform. Instead of context-switching between your IDE and the ServiceNow portal, you can now orchestrate incidents, change requests, and approvals entirely through natural language prompts in your agent.

This is particularly useful for ops teams building automation workflows or developers who need to integrate service desk operations into their deployment pipelines. The server speaks both directions—your AI agent can read incident data, search the knowledge base, and write changes back to ServiceNow in real time.

What It Does

ServiceNow MCP unlocks full ITSM lifecycle management through your AI agent. You can create and update incidents, problems, and change requests; search knowledge base articles to surface solutions during troubleshooting; manage approval workflows and SLAs; query asset and configuration item data; and access service desk analytics to identify trends.

The server acts as a direct API bridge, so your agent can handle repetitive ticket routing, auto-assign based on skill tags, fetch related configuration items during incident triage, or even generate reports from trend data. Think of it as giving Claude or Cursor read-write access to your entire service desk—without the browser tab.

How to Install

Run this single command:

npx -y @servicenow/mcp

Then add the server to your Claude Desktop configuration (or Cursor/Windsurf equivalent):

{
  "mcpServers": {
    "servicenow-mcp": {
      "command": "npx -y @servicenow/mcp",
      "env": {
        "SERVICENOW_INSTANCE": "your-instance.service-now.com",
        "SERVICENOW_API_KEY": "your-api-key"
      }
    }
  }
}

Replace SERVICENOW_INSTANCE and SERVICENOW_API_KEY with your credentials. See the official docs for authentication details.

Real-World Use Cases

Incident triage in Cursor: Paste an error log, ask Claude to search the knowledge base for solutions, and auto-create an incident with relevant config items attached—all without leaving your editor.
Change request automation: Describe a deployment, let your AI draft a change request with SLA calculations, and route it to the approval queue based on impact level.
On-call escalation: Build a Slack bot or CLI tool that queries open incidents, fetches analytics, and drafts summaries for your incident commander—all via natural language.

Full install guides for Claude Desktop, Cursor, Windsurf, and more at CuratedMCP.

MCP Ecosystem Week 19: Data Intelligence and Safety Tools Lead New Arrivals

curatedmcp — Mon, 04 May 2026 10:38:54 +0000

MCP Ecosystem Week 19: Data Intelligence and Safety Tools Lead New Arrivals

The MCP marketplace continues to expand with four new servers landing this week, reflecting a clear market demand for data accessibility and content safety. Week 19 brings tools spanning SEO analytics, conversational UI, marketing data aggregation, and threat detection—a diverse mix that signals the ecosystem is moving beyond developer-centric integrations toward enterprise and consumer-facing use cases.

This Week in MCP

Four new verified servers joined the catalog:

DataForSEO MCP Server opens SEO and SERP data to any MCP client. For developers building competitive analysis or content tools, this eliminates the friction of API documentation and token management.

Perspective AI replaces form-based interactions with adaptive AI conversations. The shift from static inputs to natural dialogue is gaining traction—this server makes that pattern pluggable for any agent.

Windsor.ai MCP Server aggregates marketing, sales, and customer data across 325+ sources and exposes it via plain English queries. This is a meaningful consolidation play: instead of agents juggling multiple vendor SDKs, they get a unified interface.

Tuteliq detects grooming, bullying, fraud, and 16+ threat types across text, voice, image, and video. Content moderation and safety are increasingly non-negotiable for production AI systems, and multimodal threat detection filling a real gap.

All four are available free—a continued pattern in the ecosystem.

Trending Servers

The top three trending servers by page views this week reveal what developers are actively integrating:

GitHub Copilot MCP (98K views) leads. The appeal is clear: developers want Copilot's code intelligence—completions, explanations, reviews—available outside the IDE, in their own MCP workflows. It's a portability play.

OpenAI MCP (87K views) maintains second place. Official integrations matter. Having GPT-4o, DALL-E, Whisper, and Embeddings available as MCP resources means developers can route AI workloads through standardized tooling rather than managing OpenAI SDK fragmentation.

Figma MCP (82K views) rounds out the top three. Design-to-code and design system integration are high-friction problems. Direct access to design files, components, and tokens within an AI agent workflow is a legitimate productivity unlock—and the traffic reflects that need is real.

The pattern: developers prioritize integration servers that reduce context switching and vendor lock-in.

Quick Take

The all-free pricing on this week's new arrivals (and 67 of 68 total servers) reflects a strategic choice by vendors: expand adoption fast, own the integration layer, monetize upstream or at the platform level. But there's a secondary story: we're seeing category clustering around data intelligence (SEO, marketing, CRM aggregation) and safety/moderation. These are problems that individual AI developers can't easily solve in-house. The marketplace is becoming a repository of specialized, repeatable capabilities—the opposite of monolithic platforms. That's healthy ecosystem behavior.

Browse all 68 verified servers at CuratedMCP.

Okta MCP: Let Your AI Agent Handle Identity Management

curatedmcp — Sat, 02 May 2026 10:12:28 +0000

Okta MCP: Let Your AI Agent Handle Identity Management

Identity and access management shouldn't require manual ticket shuffling. Okta MCP is the official Okta integration for AI agents—it lets Claude, Cursor, or Windsurf directly manage users, groups, authentication workflows, and security events within your Okta tenant.

If you're managing identity infrastructure, this means your AI can provision users, enforce access policies, investigate authentication failures, and respond to security incidents without you leaving your IDE or chat window.

What It Does

Okta MCP grants AI agents safe, scoped access to your identity workflows. Your agent can:

Manage users at scale: create, update, deactivate accounts, reset passwords, and manage recovery rules
Control access: assign users to groups and applications, manage roles, and inspect sign-on policies
Monitor security: query authentication events, check MFA status, and analyze access patterns
Automate OAuth/OIDC: manage client configurations and token policies

The server is built for security and IT teams who want AI to reduce toil—faster user onboarding, quicker incident response, and easier policy enforcement. Everything is gated by your Okta API token, so permissions stay locked to what you've explicitly authorized.

How to Install

npx -y @okta/mcp

Then add it to your Claude Desktop config:

{
  "mcpServers": {
    "okta-mcp": {
      "command": "npx -y @okta/mcp",
      "env": {
        "OKTA_DOMAIN": "your-org.okta.com",
        "OKTA_API_TOKEN": "your_api_token_here"
      }
    }
  }
}

Grab your API token from Admin → API → Tokens in your Okta console. Full setup guides for Cursor and Windsurf are in the docs.

Real-World Use Cases

Onboarding sprint: Ask Claude to provision 12 new contractors, assign them to the engineering group, activate their Slack app, and send a summary—all in one prompt.
Incident response: When a security alert fires, ask your agent to pull MFA status for a suspicious user, review their recent sign-ons, and deactivate sessions—no context switching.
Access review automation: Have your agent query who has admin roles, who hasn't logged in this quarter, and which legacy OAuth clients are still active—then generate a compliance report.

Full install guides for Claude Desktop, Cursor, Windsurf, and more at CuratedMCP.

IBM Cloud MCP: Control Your IBM Cloud Infrastructure From Claude

curatedmcp — Fri, 01 May 2026 10:06:17 +0000

IBM Cloud MCP: Control Your IBM Cloud Infrastructure From Claude

Managing IBM Cloud resources usually means jumping between dashboards, CLIs, and documentation. IBM Cloud MCP changes that by giving AI agents direct access to your infrastructure, databases, and AI services. Built by IBM for enterprise teams, this official server lets you provision compute, query data, and inspect security policies without leaving your Claude chat or code editor.

What It Does

IBM Cloud MCP is a Model Context Protocol server that bridges Claude, Cursor, and other AI agents to IBM Cloud's full platform. Instead of manual API calls or console clicks, you can ask Claude to spin up Kubernetes clusters, deploy serverless functions, or query Watson AI models directly.

The server handles compute provisioning (Kubernetes, VMs, Cloud Functions), data operations (Cloudant, Db2, Databases for PostgreSQL), and enterprise workflows like identity management and security audits. You get real-time access to logs, monitoring alerts, and metrics across regions. It's designed for teams managing hybrid cloud workloads—IBM Cloud alongside on-premises systems.

For developers, this means faster iteration on cloud-native apps, faster infrastructure troubleshooting, and the ability to automate repetitive cloud tasks through natural conversation.

How to Install

Install via npm:

npx -y @ibm/mcp

Then add it to your Claude Desktop config file (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "ibm-cloud-mcp": {
      "command": "npx -y @ibm/mcp",
      "env": {
        "IBM_CLOUD_API_KEY": "your-api-key-here"
      }
    }
  }
}

Grab your IBM Cloud API key from the IAM dashboard, set it as an environment variable, and restart Claude.

Real-World Use Cases

Deploy and scale an app in seconds: "Deploy my Node.js app to IBM Cloud Kubernetes with 3 replicas across Dallas and London regions, then show me the logs."
Debug database issues with context: "Query my Db2 instance for slow queries in the last hour, then suggest indexes based on the execution plans."
Audit security posture: "List all IAM policies for my service accounts, identify overly permissive roles, and generate a compliance report."

Full install guides for Claude Desktop, Cursor, Windsurf, and more at CuratedMCP.

Adobe MCP: Connect Claude to Photoshop, Acrobat, and Experience Cloud

curatedmcp — Thu, 30 Apr 2026 10:06:16 +0000

Adobe MCP: Connect Claude to Photoshop, Acrobat, and Experience Cloud

Adobe just released an official MCP server that plugs your AI agent directly into Creative Cloud and Experience Cloud. If you're building workflows that need to generate assets, manage PDFs, or pull marketing analytics—this cuts out the manual handoff.

What It Does

Adobe MCP gives Claude, Cursor, and other AI agents direct access to Adobe's creative and marketing tools. You can now:

Create and edit images in Photoshop programmatically—useful if you're generating variations, applying filters, or building batch workflows
Generate and manage PDFs via Acrobat—ideal for automated document creation, form filling, or report generation
Query Experience Cloud for campaign and analytics data—pull reporting directly into your AI context
Manage Creative Cloud assets and access libraries, templates, and brand-approved styles
Export finished work with delivery URLs—no manual export steps

This is officially supported by Adobe, so you're integrating with their actual APIs, not scraping or reverse-engineering. The server runs on your machine or server and communicates with your Adobe account.

How to Install

npx -y @adobe/mcp

Then configure it in your Claude Desktop config (or Cursor/Windsurf):

{
  "mcpServers": {
    "adobe-mcp": {
      "command": "npx -y @adobe/mcp"
    }
  }
}

You'll need Adobe API credentials (client ID and secret from the Adobe Developer Console). Full setup docs at developer.adobe.com/mcp.

Real-World Use Cases

Marketing automation: Ask Claude to pull last month's campaign performance from Experience Cloud, then generate a summarized PDF report with embedded charts—all in one prompt.
Batch creative generation: Feed Claude a product catalog and ask it to generate hero images for each item in Photoshop, apply your brand template, and export URLs for your website.
Document workflows: Automate contract or invoice generation by having Claude populate Acrobat templates with data, then export signed PDFs ready for archival.

Full install guides for Claude Desktop, Cursor, Windsurf, and more at CuratedMCP.

Oracle Cloud MCP: Automate OCI infrastructure through AI agents

curatedmcp — Wed, 29 Apr 2026 10:06:16 +0000

Oracle Cloud MCP: Automate OCI infrastructure through AI agents

Oracle Cloud MCP is the official integration that connects AI agents like Claude and Cursor directly to Oracle Cloud Infrastructure (OCI). Instead of switching between terminals and the OCI console, you can now ask your AI assistant to provision instances, manage databases, handle storage, and configure security—all within your IDE or chat interface.

What It Does

This server bridges the gap between AI agents and OCI operations. It gives Claude, Cursor, and Windsurf read and write access to your OCI environment through the Model Context Protocol. Your AI can:

Compute: Spin up, terminate, and monitor compute instances with different shapes and configurations
Storage: Create and manage Object Storage buckets, upload objects, and handle lifecycle policies
Databases: Query Autonomous Database instances, manage DB systems, and inspect schemas
Networking: Inspect Virtual Cloud Networks (VCNs), review security lists, and audit firewall rules
Identity & Access: Create IAM users, attach policies, and manage permissions programmatically
Security & Compliance: Pull audit logs and generate security reports for governance

It's built for enterprise teams who run OCI and want to shift repetitive infrastructure tasks to AI automation.

How to Install

Install via npm:

npx -y @oracle/mcp

Then add it to your Claude Desktop configuration:

{
  "mcpServers": {
    "oracle-cloud-mcp": {
      "command": "npx -y @oracle/mcp",
      "env": {
        "OCI_CONFIG_FILE": "~/.oci/config",
        "OCI_PROFILE": "DEFAULT"
      }
    }
  }
}

Make sure your OCI credentials are configured locally before connecting.

Real-World Use Cases

On-demand infrastructure scaling: Ask Claude to check current compute usage, then provision additional instances if load is high—all without leaving your editor
Database troubleshooting: Query Autonomous Database logs, inspect connection pools, and get optimization suggestions from your AI in real time
Audit and compliance: Generate security reports, review IAM policies, and pull audit logs to validate your infrastructure against compliance requirements

Full install guides for Claude Desktop, Cursor, Windsurf, and more at CuratedMCP.

Elastic Cloud MCP: Give Claude Direct Access to Your Elasticsearch and Kibana Data

curatedmcp — Tue, 28 Apr 2026 10:06:17 +0000

Elastic Cloud MCP: Give Claude Direct Access to Your Elasticsearch and Kibana Data

Searching logs manually is slow. Analyzing metrics across dashboards takes context switching. What if Claude could do it for you?

Elastic Cloud MCP is the official integration that plugs your Elasticsearch and Kibana deployment directly into Claude, Cursor, and other AI agents. Instead of copying query syntax or explaining your schema, you can ask Claude to investigate anomalies, explore logs, or run analytics—and it talks to your Elastic Cloud instance in real time.

Built by Elastic and verified for production use, this server gives AI agents safe, structured access to your search and observability stack without exposing raw credentials or requiring manual data extraction.

What It Does

Elastic Cloud MCP turns your Elasticsearch indexes and Kibana dashboards into tools Claude can use directly.

You can ask Claude to:

Query indexes and run analytics across your data in natural language
Inspect dashboards and visualizations to understand what's already built
Search logs, metrics, and traces without switching tools
Explore schema mappings and see which fields are available
Set up alerts and anomaly detection jobs programmatically
Manage index templates and ingest pipelines for data ingestion flows

This is especially powerful for incident response—Claude can query your logs, correlate metrics, and surface the context you need without you leaving your editor or chat window.

How to Install

Install via npm:

npx -y @elastic/mcp

Then configure it in your claude_desktop_config.json:

{
  "mcpServers": {
    "elastic-cloud-mcp": {
      "command": "npx -y @elastic/mcp",
      "env": {
        "ELASTIC_CLOUD_ID": "your-deployment-id",
        "ELASTIC_API_KEY": "your-api-key"
      }
    }
  }
}

Get your Cloud ID and API key from your Elastic Cloud console, then restart Claude Desktop.

Real-World Use Cases

Incident triage: Ask Claude to search error logs from the last hour, pull relevant metrics, and summarize what went wrong—all without manual queries.
Schema exploration: New to a team's data model? Have Claude map out available fields, data types, and common usage patterns from your live indexes.
Dashboard-driven analysis: Point Claude at a Kibana dashboard and ask it to explain the data, suggest optimizations, or dig deeper into anomalies it finds.

Full install guides for Claude Desktop, Cursor, Windsurf, and more at CuratedMCP.

MCP Ecosystem Week 18: The Consolidation Phase — When Integration Depth Beats New Tools

curatedmcp — Mon, 27 Apr 2026 10:57:56 +0000

MCP Ecosystem Week 18: The Consolidation Phase — When Integration Depth Beats New Tools

The MCP marketplace hit an interesting inflection point this week. For the first time in our roundup series, we saw zero new server additions — yet engagement across the catalog remained strong. This isn't a slowdown. It's a signal that the ecosystem is shifting from breadth to depth, with developers and AI agents diving deeper into fewer, more powerful integrations rather than constantly chasing new tools.

With 64 verified servers now in the catalog and 100% of them free to use, the market is consolidating around proven, official integrations and specialized use cases. The data tells a clear story: when you can do more with fewer tools, you do.

This Week in MCP

No new servers entered the verified catalog this week, marking a notable pause in new submissions. This doesn't reflect reduced activity — rather, it suggests that recent additions from previous weeks are still onboarding, being tested, and integrated into workflows.

The verification backlog remains healthy, and community submissions continue. If you've built an MCP server and want to list it, the team is actively reviewing submissions at CuratedMCP.

Trending Servers

1. GitHub Copilot MCP (98,002 views) sits firmly at the top, offering direct access to Copilot's code intelligence — completions, explanations, and automated review — from any MCP client. For developers working across multiple editors and AI agents, this is table stakes. It's the connective tissue between where you write code and where AI helps you write it better.

2. OpenAI MCP (87,001 views) provides official access to GPT-4o, DALL-E, Whisper, and embeddings. What's driving sustained interest here is flexibility: rather than being locked into OpenAI's native interfaces, developers can compose these capabilities into custom workflows. That's the core MCP value proposition in action.

3. Figma MCP (82,003 views) rounds out the top three, letting design systems and component libraries flow directly into AI coding workflows. This one is particularly interesting because it bridges a gap that rarely gets bridged — design metadata meeting code generation. Teams using component-driven architecture are clearly building around this integration.

All three trending servers are official integrations from major platforms (GitHub, OpenAI, Figma). That pattern matters.

Quick Take

The "all free" catalog is a competitive advantage, not a limitation. With zero paid servers in the verified directory, CuratedMCP has become the canonical destination for developers who want production-grade integrations without licensing complexity.

More importantly: the dominance of official platform integrations (GitHub, OpenAI, Figma, Anthropic Claude all in the top five) suggests the MCP ecosystem is following a predictable adoption curve. Phase one was experimental diversity. Phase two — where we are now — is consolidation around the integrations that unlock multiplier effects: code, design, AI reasoning, and version control, all composable in one workflow.

Watch for phase three: specialized vertical tools (domain-specific MCPs for legal tech, biotech, finance) layering on top of this foundation. That's where real differentiation emerges.

Browse all 64 verified servers at CuratedMCP.

MongoDB MCP: Give Your AI Agent Direct Access to Your MongoDB Clusters

curatedmcp — Mon, 27 Apr 2026 10:57:49 +0000

MongoDB MCP: Give Your AI Agent Direct Access to Your MongoDB Clusters

MongoDB just released an official MCP server, and it's a game-changer for developers who want their AI agents to actually work with NoSQL data instead of just talking about it.

What It Does

MongoDB MCP lets Claude, Cursor, and other AI agents query, inspect, and modify MongoDB Atlas clusters and local instances directly. Instead of asking an AI to write MongoDB queries and hoping they're correct, the agent can now execute queries, explore your schema in real time, and handle document operations autonomously.

The server supports the full CRUD lifecycle: query documents with filters and aggregations, inspect collections and indexes, create and update records, and manage cluster configuration. You also get performance monitoring, slow query detection, Atlas search integration, and change stream support — everything a developer needs to orchestrate database workflows through natural language.

This is especially powerful for exploratory work. Ask your AI to "find all documents where status is 'pending' and createdAt is older than 30 days" and it executes the actual query, sees the results, and can act on them. No more back-and-forth copy-pasting.

How to Install

Install via npm:

npx -y @mongodb/mcp

Then add to your Claude Desktop config file (claude_desktop_config.json):

{
  "mcpServers": {
    "mongodb-mcp": {
      "command": "npx -y @mongodb/mcp"
    }
  }
}

You'll need MongoDB connection credentials (Atlas connection string or local instance details). Full setup instructions are in the MongoDB MCP docs.

Real-World Use Cases

Data migration and cleanup: Ask Claude to find and transform legacy documents across collections, then execute the updates. No manual scripting required.
Schema exploration for feature development: Query your actual data patterns, discover edge cases in your schema, and get AI suggestions for indexing improvements based on real access patterns.
Debugging production issues: Have your AI agent query slow logs, inspect problematic documents, and suggest fixes — all without leaving your chat interface.

It's free and officially maintained by MongoDB, so you're not relying on a community fork.

Full install guides for Claude Desktop, Cursor, Windsurf, and more at CuratedMCP.