Pricing is per user, per month, starting at ten dollars for each person connected. The rate comes down as your team grows. You pay for the people using it rather than for the number of connectors you attach.

Managed MCP Gateway for AI Teams | Cascadia Web Services

Cascadia engineer writing WordPress code across dual monitors.

Built for teams running many tools

What makes this MCP gateway different

Most teams wire each MCP server into each assistant by hand. That works fine for one person with two tools. It stops working the moment you have five people and ten providers, because every combination is its own piece of setup and its own credential sitting in somebody's config file. A gateway changes the shape of that problem. You attach a provider once, then decide who reaches what.

One endpoint, every connector

Attach as many providers as you like and they all arrive through the same connection. Your assistant sees a single tool list covering everything you have been granted, whether a given connector runs inside your own instance or somewhere else entirely.

Access decided per person

Enable and disable individual tools for each user. Two people can share the same gateway and see completely different tool lists, and neither of them has to configure anything differently for that to happen. You decide the shape of it centrally.

Nothing for you to run

We host the gateway and keep it running. There is no container to deploy, no process to babysit, and no certificate to renew before it expires. Sign up, attach your providers, and point your assistant at one address.

Credentials never leave the server

Provider keys stay on the gateway instead of being copied into every person's assistant configuration. When one needs rotating you change it in a single place, and everybody connected keeps working, because none of them were holding that key to begin with.

Change access without reconnecting

Adjust what somebody can reach and their existing connection picks it up. No reconnecting, no re-authenticating, and no asking a colleague to edit a file on their laptop first. Granting and revoking become administrative actions instead of support tickets.

One place to revoke

When somebody leaves, you remove their access once. Because everyone connects through the same endpoint instead of holding their own direct connections, there is no hunt through separate configurations to find every server they were ever handed.

Gateway versus direct

An MCP gateway vs wiring each server by hand

Connecting servers directly is perfectly fine at one person and two tools. The cost shows up as you add either. Every person multiplied by every server is a separate configuration entry to create, document, and later remember to remove. These are the six places that difference actually bites.

One connection vs one per server

Your assistant holds a single entry pointing at the gateway. Connecting directly means a separate entry for every server, recreated by hand on every machine and in every client that needs to reach it.

Access per tool vs all or nothing

Grant somebody three tools out of a connector and leave the rest closed. A direct connection hands over whatever that server exposes, because there is no layer in between with the authority to narrow it down.

Keys held once vs pasted everywhere

The gateway stores provider credentials centrally. Direct connections put a copy of every key into every person's configuration file, which is where credentials go to leak and where rotation quietly fails to reach.

Add once vs set up for everyone

Attach a new provider and grant it to whoever needs it. Without a gateway, each person repeats that setup themselves, and you find out who missed a step when something quietly does not work for them.

Revoke once vs chase every config

Removing somebody is one action. With direct connections you have to know every server they held and reach every machine storing those credentials, which in practice means some access quietly survives.

One tool list vs a pile of entries

Everything you have been granted arrives together, tagged by the connector it came from. Managing a dozen separate entries per client is where setup drifts and people end up running different versions of it.

Getting started

How setting it up actually goes

Three steps, and none of them involve installing software. You create an account, attach the providers you want to reach, and add a single entry to your assistant. Granting other people access happens after that, and takes about as long as ticking boxes.

Sign up and create your gateway

Signing up is free and takes a minute. You get your own gateway endpoint, which is the single address every assistant and every person on your team points at from here on. Nothing is installed on your side, there is no infrastructure to provision first, and you can attach your first provider straight away.

Attach the providers you want

Add each MCP server you want to reach, whether it is one of ours, one you already run yourself, or a third party service. Credentials are stored once, on the gateway, rather than handed out to everybody who needs the tools behind them. Attach as many as you like now, or come back and add more later.

Grant people the tools they need

Add your team and choose, tool by tool, what each person can reach. They add one entry pointing at the same address you use, and see exactly what you granted and nothing else. Change somebody's access later and their existing connection picks it up, so nobody has to reconnect or edit anything themselves.

Testimonials

Here's what others had to say

Our team handles daily updates to the WordPress core, ensuring your site is always running the latest and most secure version. These updates include performance tweaks and database optimizations, helping your site stay fast and stable. We also clear cached data and proactively fix issues before they affect users. With us managing your updates, you never have to worry about falling behind or breaking your site.

Joe Q.

I can't say enough about how grateful I am to [Cascadia] for helping me resolve my tech problems. I was in a real bind, and [they] calmly and cooly fixed the problem--something two other tech support folks could not do. [They are] gonna be my go-to from now on.

Sandi S.

Naomi T.

[Cascadia] is amazing! They are so patient and explains things in such a clear way. I'm very grateful to them for making me feel more confident in my work with the CRM. Can't recommend them enough!!

Aventurina K.

I've been on Upwrok for 5+ years hiring various contractors for my service based business and there is no doubt that [Cascadia] is one of the top 5 I've ever hired for ANYTHING. For domain related questions, a selfless work-ethic, and a deep & reliable source of knowledge, [Cascadia] is the right choice.

Andrew K.

I've worked with Cascadia for nearly three years, and the service is fantastic. The benefits are reasonably priced, and the turnaround time is better than any titan within the industry. I love this company, and we look forward to growing with them.

Carl B.

Everything included

What the router includes

All

One Endpoint For Everything

Point your assistant at a single URL and reach every connector you have enabled. Adding a provider no longer means adding another connection to every client that needs it.

Attach Unlimited Connectors

There is no cap on how many providers you attach. Each one joins the same endpoint, and the tools it exposes appear alongside everything else already connected.

Local And Remote Providers

Attach connectors running inside your own instance and connectors reachable over HTTPS at the same time. The assistant sees one list of tools and does not care where each one lives.

Live Connection Reporting

Ask the router what it is connected to and it answers with the client identity, every active integration and its address, and the total number of tools currently exposed.

Tools Stay Namespaced

Every tool arrives tagged with the connector it came from, so two providers offering similarly named operations never collide or get confused for one another.

Per User, Per Tool Access

Enable and disable individual tools for each person. Two people can share the same router and see completely different tool lists, without either of them configuring anything themselves.

One Connection, Scoped Centrally

Each person connects once to the same endpoint. What they can actually reach is decided centrally, so an access change takes effect without anyone editing a configuration file.

Grant By Provider Or By Tool

Turn on an entire connector for someone, or pick out individual tools from it. Access is shaped to the person rather than handed over in one block.

Change Access Without Reconnecting

Adjust what somebody can reach and their existing connection picks it up. They do not need to reconnect, re-authenticate, or touch their assistant configuration for it to apply.

Revoke In One Place

Removing a person's access happens once, centrally. There is no hunt through separate client configurations to track down every connection they were given along the way.

Central Configuration

Provider addresses, credentials, and settings live in one place instead of being copied into every assistant that needs them and drifting out of sync over time.

Add A Provider Once

Attach a new connector and it becomes available to everyone you grant it to. There is no per-user setup step and no credential to reissue to each person.

Managed And Hosted

The router runs on our infrastructure. There is nothing for you to deploy, patch, monitor, or keep running, and no server of your own sitting in the path.

Named Client Identity

Every connection carries the identity of the person using it, so activity through the router is attributable to a named individual rather than to one shared anonymous key.

Token Authenticated Connectors

Downstream connectors accept traffic only when it carries the shared secret the router holds, so a publicly reachable connector rejects anything arriving without it.

Credentials Stay Server Side

Provider keys are held by the router rather than distributed into each person's assistant configuration, where they are easy to leak and painful to rotate afterwards.

Answers to common questions about the router

Frequently asked questions

What is the MountDev AI MCP Router?

It is a managed MCP gateway. Instead of wiring each MCP server into each AI assistant separately, you attach your connectors to the router once and point your assistant at a single endpoint. Everything you have been granted shows up in one tool list. We host and run it, so there is no server for you to deploy or maintain.

How is this different from connecting each MCP server directly?

Connecting directly means one configuration entry per server, per client, per person. Ten tools across five people is fifty pieces of setup to create and later maintain. The router collapses that to one connection each. You attach a provider once centrally, and everyone you grant it to picks it up without touching their own configuration.

Is this an MCP gateway?

Yes. Gateway, router, proxy, and aggregator all describe the same pattern: a single service sitting between your AI assistants and the MCP servers behind them. We call ours a router because it also decides which person reaches which tool, rather than only forwarding traffic to whatever sits downstream.

How many connectors can I attach?

There is no limit. Each connector you attach joins the same endpoint, and its tools appear alongside everything else already connected. Tools stay tagged with the connector they came from, so two providers offering similarly named operations do not collide or get mistaken for one another.

How do per-user permissions work?

Access is set per person and per tool. You can enable a whole connector for somebody, or pick out individual tools from it. Changes apply to their existing connection, so nobody has to reconnect, re-authenticate, or edit a config file when you adjust what they can reach.

Can two people share one router and see different tools?

Yes, and this is the main reason to run one. Both people connect to the same endpoint with the same setup steps, but each sees only the tools you have granted them. Their assistants show different tool lists without either person configuring anything differently.

Where does the router run?

On our infrastructure. It is a managed service, so you connect to an endpoint we operate rather than standing anything up yourself. That means no deployment, no patching, and no uptime to watch. If you would rather self-host the individual connectors, our Frappe and WordPress connectors are free and open source for exactly that.

Do I need to install or host anything?

No. You sign up, attach the providers you want, and add one entry to your assistant's configuration pointing at the router. There is no package to install, no container to run, and no certificate to manage. The connectors behind the router can be ones we host or ones you already run yourself.

What does it cost?

Pricing is per user, per month, starting at ten dollars for each person connected. The rate comes down as your team grows, so larger groups pay less per seat. You pay for the people using it rather than for the number of connectors you attach or the tools they reach, which means adding another provider never changes your bill.

Is there a free option?

Yes. A free account gives you one user with no limit on how many connectors you attach, so you can wire up your real providers and use it properly rather than poking at a demo. Adding a second person is what moves you onto paid seats, at which point the per user rate applies across the account.

What can I connect to it?

Any MCP server the router can reach. That includes connectors running locally inside an instance and connectors reachable over HTTPS, at the same time, in the same tool list. Our own Frappe and WordPress connectors are built to sit behind it, and any standards compliant MCP server works the same way.

How are provider credentials handled?

The router holds them. Keys stay server side rather than being copied into each person's assistant configuration, where they are easy to leak and painful to rotate. When a credential changes you update it once centrally, and every connected user keeps working because none of them were holding it.

How do I remove someone's access?

In one place. Because everybody connects through the same endpoint rather than holding their own direct connections, revoking a person is a single action rather than a hunt through separate client configurations to find every connector they were ever given.

Can I use it with Claude Desktop?

Yes. Add the router as a connector in your client and it appears as one entry exposing every tool you have been granted. This works the same way in any MCP compatible client, and it replaces the separate entry you would otherwise need for each individual server.

How do I see what is currently connected?

Ask the router. It reports back the client identity it sees, every active integration with its address, and the total number of tools currently exposed to you. That gives you a quick way to confirm a new provider attached correctly and that a person is seeing what you expect them to see.

One MCP Gateway For All Your Connectors Tools Users