mcp-use is the open-source devtools and cloud infrastructure to help dev teams quickly build and deploy custom AI agents with MCP servers. Our SDK has over 5,000 GitHub stars, 100k downloads, and is trusted by engineers at NASA, Cisco, NVIDIA, etc...












mcp-use
Hi Product Hunters 🐱👋
I’m Luigi, cofounder of mcp-use (YC S25) with @pietrozullo .
mcp-use is the open-source dev tools and infrastructure for MCP to help dev teams quickly build and deploy custom AI agents with MCP servers.
mcp-use SDK just crossed 120,000 downloads and 6,000 GitHub stars.
Dev teams use us at a variety of companies, both startups and enterprises like NASA, NVIDIA, SAP, to build agentic products or internal custom agents with MCP.
🚨 The problem we’re solving
Correctly build and deploy MCP servers.
Fragmented MCP server configs (in GitHub repos, hardcoded in code, random registries).
Handle auth, access control, and audit logging.
Reduce the number of tools exposed (too many tools → LLMs get confused).
Manage environments.
MCP agents running locally (not in a safe, observable runtime).
Observability gap.
🥁 What you get
Dev teams can build MCP agents in their products using mcp-use SDK, which is integrated with mcp-use Cloud Platform, the central control plane layer that acts like a gateway for all the MCP servers.
mcp-use provides a vertical solution for MCP development:
1️⃣ mcp-use SDK:
Easily integrate MCP-enabled AI agents into your product or internal tools.
2️⃣ mcp-use Cloud Platform:
The central control plane layer for MCP servers, managing configs, server selection, caching, metrics, and access control.
3️⃣ mcp-use Server Hosting:
Managed/self-hosted servers, third-party MCP servers, and short-lived stdio sandboxed servers.
🏰 Our story
When MCP and the first use cases came out, we could not believe that such powerful tech could only be used on IDEs or Claude Desktop.
We felt the need to write agents ourselves easily, in code, and allow other developers to do the same.
That’s why we first released mcp-use library.
Now we want to make the development of MCP agents dev-friendly and production-ready.
❓ Why mcp-use?
Our cloud platform provides developers with a single, unified interface for MCP.
You can configure multiple MCP servers into a single pool, creating agents tailored to your applications.
You can integrate these agents through mcp-use SDK with just one line of code and embed them into your products.
We handle all the hosting and deployment complexities.💆
Fast setup: connect many MCP servers into one pool.
Safer by default: profile-based access, audit logs, tool limits.
Prod-ready: environments, rollbacks, observability, and a sandboxed agent runtime.
Think Vercel + Next.js, but for MCP.
🚀 Special launch offer
Try it out with personalized onboarding.
Start here 👉 https://mcp-use.com/
Libraries 👉 Python: https://github.com/mcp-use/mcp-use | Typescript: https://github.com/mcp-use/mcp-u...
We created mcp-use to be the vertical solution for MCP.
What do you think?
P.S. Drop by the comments, we’d ❤️ your feedback!
👇
HookWatch
The gap between "MCP works in my IDE" and "MCP works in production" is massive, and this tackles it head on. The managed authentication piece is what enterprises are blocking on — nobody wants to give an agent raw credentials without proper access control. One thing I keep running into: once agents are deployed, the observability story falls apart. You can build the workflow, but when an MCP tool call fails at 2am, there's no structured way to trace what happened, what payload was sent, and what the server returned. Are you thinking about building a monitoring/logging layer into the platform, or do you see that as a separate concern?
@pederzh @pietrozullo congrats on the launch, have upvoted. Do you have a Miro MCP that works on Claude? Also how do you protect the mcps from code injection risk?
mcp-use
@shashwat_ghosh_gtm a Miro MCP server exists, tbh I've never used it. You can add it on mcp-use cloud (https://mcp-use.com) by just copying and pasting the server configuration and you can easily spin up an MCP agent that uses it on the web that can be accessed by any device.
@pederzh I'd seen that but its not officially endorsed by Claude or Miro...so is it safe to use?
mcp-use
@shashwat_ghosh_gtm yes
Congrats on the momentum! Clean OSS SDK with cloud control plane is a strong combo for MCP teams.
I think PQL is, first agent registered → first request routed through the control plane → first managed server live (or first shadow bill pass), right?
I have few suggestions, if you lean towards billing/metering. Let me know what you think!
• Run 1–2 shadow cycles before going live - so request counts and server hours match what customers expect on invoices.
• Add a gentle in‑app usage meter with 50/80/100% alerts & a one‑click "lock spend" to prevent surprises and keep trust high.
• Consider hybrid packaging - seats for admin/collab + usage for requests/servers with a small monthly minimum and volume breaks. Predictable for you, flexible for them.
• Grandfather existing OSS users who migrate to the cloud for a reasonable window to smooth adoption.
Congrats on the launch 🚀 sounds like you’re tackling the same pain point everyone hits after the first cool demo: configs scattered everywhere, security duct-taped on, and zero observability.
Turning MCP into something as approachable as Vercel made web dev is a big deal. Curious, how are you thinking about scaling policies (access, audit, rollbacks) across teams at enterprise level?
Pally - AI Relationship Management
let’s goooo! was great to see an earlier version of this - the managed authentication for agents and access is HUGE for enterprises. congrats on the launch!
mcp-use
@hazhubble thanks! Absolutely, MCP has gained great adoption within companies, but what we've seen so far is that the rollout is often blocked due to missing tools for authentication, access control, and observability.
In fact, we provide an MCP gateway where all the requests pass through, and we are able to plug all these services!
panda{·}etl (YC W24)
Congrats for the launch, great product and great team! @pederzh
mcp-use
@gabriele_venturi thanks Gabri! Your compliment is especially important coming from you.