Build AI agents with 600+ integrations in hours, not months. Metorial's MCP-powered platform handles OAuth, monitoring, and deployment automatically. Open-source with Python & TypeScript SDKs. Easily self-hostable.
Before building Metorial, my co-founder (Wen) and I were working on a drag-and-drop AI agent builder. We found that MCP works great for client-side tools like Cursor and Claude Desktop, but integrating MCP servers on the backend for cloud-hosted AI agents is a mess. Managing infrastructure, handling OAuth, dealing with cold starts, losing state between connections, and getting everything to work reliably.
We built Metorial to fix that.
What Metorial does: We make it really easy for developers to connect their AI agents to 600+ integrations ranging from Slack and Gmail to Salesforce and Hubspot. We give you really nice SDKs and APIs to connect those integrations to your agents. We run any MCP server on truly serverless infrastructure. Not "serverless" where your server dies and loses everything. We can pause and resume without dropping state or interrupting connections. We're the only ones doing that.
What you get:
600+ MCP integrations - Connect your AI agents to 600+ services via MCP (Google Calendar, Sentry, Salesforce, GitHub, etc.)
Truly serverless - We're the only platform that can pause and resume any MCP server without losing state or causing interruptions. MCP sessions with Metorial can live for week and only incur seconds of compute time
One-line OAuth - We handle OAuth for you in literally one line of code
Full observability - Every message, tool call, and error is traceable in our dashboard
Built for devs - Python and TypeScript SDKs and an API that get you up and running in minutes
Deploy, fork, customize - Use existing MCP servers or deploy your own to our infra
Versioning and sandboxes - Test your integrations first then deploy to production with our built-in versioning
The goal is simple: we help you connect your AI agents to any tool and datasource. We make it as easy to use MCP on the server side as it is on the client side. Now any AI agent can integrate with hundreds of tools in a few lines of code instead of weeks of engineering work.
Use the coupon code PH50OFF for 50% off!
Would love to hear your questions and feedback. What integrations would you want to see?
I'm planning to build just a ChatGPT app, and an MCP integration platform seems like overkill for just that one integration. But this opens up the possibilities for a ton more integrations that I didn't even think of. For example populating a google calendar.
I was planning to use LangSmith for chain/agent-level traces and alerts, but if I can get a single pane of glass just for the MCP tool calls + sessions, that alone be a big benefit, even if we just have a couple MCP integrations
Super cool work,
State persistence and observability are huge pain points, love how you’re solving that for MCP at scale. Excited to see where Metorial goes!
Very cool! I am wondering if you are thinking of supporting other apps like loops and granola? How does one go about requesting new apps too?
Hey Product Hunt!
Tobias here, founder of Metorial (YC F25).
Before building Metorial, my co-founder (Wen) and I were working on a drag-and-drop AI agent builder. We found that MCP works great for client-side tools like Cursor and Claude Desktop, but integrating MCP servers on the backend for cloud-hosted AI agents is a mess. Managing infrastructure, handling OAuth, dealing with cold starts, losing state between connections, and getting everything to work reliably.
We built Metorial to fix that.
What Metorial does: We make it really easy for developers to connect their AI agents to 600+ integrations ranging from Slack and Gmail to Salesforce and Hubspot. We give you really nice SDKs and APIs to connect those integrations to your agents. We run any MCP server on truly serverless infrastructure. Not "serverless" where your server dies and loses everything. We can pause and resume without dropping state or interrupting connections. We're the only ones doing that.
What you get:
600+ MCP integrations - Connect your AI agents to 600+ services via MCP (Google Calendar, Sentry, Salesforce, GitHub, etc.)
Truly serverless - We're the only platform that can pause and resume any MCP server without losing state or causing interruptions. MCP sessions with Metorial can live for week and only incur seconds of compute time
One-line OAuth - We handle OAuth for you in literally one line of code
Full observability - Every message, tool call, and error is traceable in our dashboard
Built for devs - Python and TypeScript SDKs and an API that get you up and running in minutes
Deploy, fork, customize - Use existing MCP servers or deploy your own to our infra
Versioning and sandboxes - Test your integrations first then deploy to production with our built-in versioning
We're open source and self-hostable: github.com/metorial/metorial
The goal is simple: we help you connect your AI agents to any tool and datasource. We make it as easy to use MCP on the server side as it is on the client side. Now any AI agent can integrate with hundreds of tools in a few lines of code instead of weeks of engineering work.
Use the coupon code PH50OFF for 50% off!
Would love to hear your questions and feedback. What integrations would you want to see?
Metorial: metorial.com
Github: github.com/metorial
Get started at: app.metorial.com
Book a demo: https://cal.com/karim-rahme/launch-yc
~ Tobias & Wen