Great questions!

If you want to run your own remote servers (for your product/company) Railway or Render work great (Vercel is a bit more difficult since Lambdas are very expensive if you run them over long periods of time). Metorial targets developers who build their own AI agents and want to connect them to integrations. Plainly, we do a lot more then running MCP servers; we give you monitoring, observability, handle consumer-facing OAuth, and give you super nice SDKs to integrate MCP servers with your agent.

Regarding the second question, Metorial has three execution modes depending on what the server supports: 1) Docker - this is the most basic one which any MCP server should support. We did some heavy optimizations to get those to start as fast as possible and our hibernation system supports stopping and resuming them while restoring the state. 2) Remote MCP - we connect to remote MCP servers for you, while still giving you the same features and ease-of-integration you get with any Metorial server (I could go more into detail on how our remote servers are better than standard ones). 3) Servers on our own lambda-based runtime. While not every MCP server supports this execution mode, it's what really sets us apart. The Lambdas only run for short intervals, while the connection is managed by our gateway. We already have about 100 lambda-based servers and working on getting more on to that execution model.

There's a lot about our platform that I haven't included in this. Like our stateful MCP proxy, our security model, our scalable SOA, and how we transform OAuth into a single REST API calls for our users.

Let me know if you have any additional questions, always happy to talk about MCP and software architecture.

Thanks so much! I'll definitely check out your book. Always happy to talk MCP :)

Haha, good thing we launched today. Thank you so much for the encouraging words!

That's a really interesting point. We've actually been discussing this quite a bit. We felt like putting an emphasize on the "dev tool" aspect (like Vercel) makes more sense, but the way you put it we might want to reconsider that. Thank for your interest!

Wen here, the co-founder. I actually spent a couple hours today to take the time to give you a comprehensive answer.

1. As you said, Composio doesn’t allow self-hosting and the source code isn’t available. We want to follow PostHog’s playbook in letting devs run everything on their own infrastructure and open sourcing all our MCP containers.

2. A huge benefit of this approach is that we can let you fork any MCP server through our dashboard so that you can manage it yourself and make any adjustments you might need. We’ve heard the importance of doing this repeatedly from our enterprise customers.

3. I do believe that we offer more robustness features, like environment provisioning, deployment versioning, server pooling, in-depth logs of server startup, as well as a complete trace of the entire MCP session.

4. On the integrations side, Composio does indeed have more integrations right now, but we already have around 600 MCP servers (all with multiple tools of course) of which many are being modified by us every day to make them better. Since we support open source contributions, the catalog also grows with the community. (Quick note that you can have private servers scoped to your org).

5. I tried to benchmark our architecture vs Composio’s in terms of speed. As we mentioned in the post above, one thing that we spent a lot of time on was optimizing how fast we can do serverless with MCP servers. However, since Composio has neither source available nor any technical documentation on how they handle their servers, I couldn’t actually find any information on their architecture. One thing that they enforce as default is having a meta-tool layer with tools like composio_search_tools and composio_execute_tool. Assuming that this is a long living process, I still found that our implementation returned a list_tools response quicker (including the cold start time). If you factor in the time that it takes for them to find the right tools, their response took close to double the time. While we might explore a similar meta tool layer as an optional MCP server in the future, we do seem to on average have a better architecture in terms of speed, though the benchmarking was not entirely rigorous. (I am also unable to answer how they handle multiple users connecting to one MCP server with different OAuth configs because they don’t share that information). I plan on making a more rigorous comparison in a blog post soon, also comparing to hosting on Vercel, Cloudflare, etc..

Let me know if you have any follow up questions.

If you want to talk more, please feel free to DM me on LinkedIn (https://www.linkedin.com/in/karim-rahme/) or X (https://x.com/wen_rahme).

Thanks for pointing that out. Ultimately, we wanted to strike a balance between being fair and open to the community, welcoming contributions, and ensuring that people can self-host without having to worry about licensing issues, while also ensuring that Metorial, as a company, can exist and work on OSS sustainably. This isn't easy and I don't think there's a right answer. To us FSL strikes a pretty good balance. Allowing the community to use and participate while ensuring that Metorial makes sense as a business as well.