MCP Hub · registry & catalog

Every tool your agents can use, in one catalog.

Connect a server once and Meandr discovers what it offers and lets you pick what your agents actually see. All your servers become one MCP registry and tool catalog — and it stays current on its own as those servers change.

Get started See how it works

All your servers, one catalog.

Connect a server once and its tools show up here — merged with the rest into a single catalog you curate, kept current on its own as those servers change.

Connected servers & catalog — screenshot placeholder
Every connected server, merged into one catalog your agents see.

Tools pile up faster than you can keep track.

GitHub here, Stripe there, a couple of internal servers — and no single place that says what your agents can actually reach.

Scattered across servers

Every server is its own island, with its own connection and its own list of tools. There's no one view of what all of them, together, let an agent do.

You can't judge a tool by its name

Some servers fold a whole API into one catch-all tool, with the real action buried in an argument. "Is this tool safe?" isn't even answerable without looking at what it's being asked to do.

The list drifts

Upstream servers add, rename, and remove tools all the time. Keep the catalog in sync by hand and it's out of date by next week.

How it works

Connect once. Curate what agents see.

Adding a server is a dashboard job, not an agent job. Meandr does the discovery and the first-pass risk read; you decide what makes the cut.

01

Connect

Point at a server.

Give Meandr a server's address and credentials. It checks it can reach and sign in to the server before treating it as live, so a broken connection shows up at setup — not mid-run.

You set up: the server endpoint and how Meandr signs in to it.

02

Discover

Meandr reads the menu.

Meandr automatically discovers the tools the server offers and keeps that up to date — on a schedule and when the server signals a change. No re-importing by hand every time something moves.

You set up: nothing — discovery and refresh are automatic.

03

Curate

You pick what ships.

You pick which tools reach your agents and set their rules. When a server declares a tool read-only or destructive, Meandr surfaces that to help you decide. Nothing reaches an agent until you've let it into the catalog.

You set up: which tools agents can see, and their starting rules.

Tool catalog with risk labels — screenshot placeholder
You decide exactly what your agents can get.

One catalog you can actually trust.

Reachable isn't the same as exposed. Meandr keeps that line clear.

One catalog, every server

All your connected servers merge into a single list your agents see as one server — and if two servers share a tool name, Meandr flags the clash for you to resolve.

A head start on the rules

When a server declares a tool as read-only or destructive, Meandr shows it and can factor it into your rules — so you're not starting from a blank page.

Reachable ≠ exposed

Connecting a server doesn't hand its tools to your agents. Nothing shows up for an agent until you've explicitly let it into the catalog.

Stays current on its own

When an upstream server changes its tools, the catalog keeps up automatically — on a schedule and on change notifications — so you're not re-importing by hand.

Connect anything — even AWS

API keys, custom headers, basic auth, OAuth, mTLS or AWS SigV4: Meandr handles the auth for you, so an AWS-hosted MCP server connects out of the box. No sidecar, no proxy to run.

Bring your stdio MCP servers

Meandr supports them. Run one on your computer or on a machine you own, and it shows up in your dashboard like any other server, ready for any of your agents.

Questions you'd ask anyway.

Does connecting a server rename the tools my agent uses?
By default, no — tools keep their original names. Renaming only happens if you choose to, or if two connected servers export a tool with the same name, in which case Meandr flags the clash and you pick the names.
Do I have to re-import every time an upstream tool changes?
No. The catalog refreshes automatically on a schedule and reacts to change signals from the server, so it keeps up without you re-importing by hand.
Does Meandr read the data behind these tools?
For the catalog, no — it only looks at tool definitions (names, descriptions, input shapes), not the data the tools operate on. The actual call content is governed separately and isn't stored unless you opt in.
What if a tool's real risk isn't obvious from its name?
Some servers declare hints — "this one is read-only," "this one is destructive" — which Meandr surfaces and can use in your rules. They're self-declared by the server, though, so treat them as a helpful signal, not a guarantee. For catch-all tools, you can also write rules on the actual arguments.
Can I connect AWS-hosted MCP servers?
Yes, out of the box. Meandr signs AWS SigV4 requests for you, so you just point it at the server — no proxy or sidecar to stand up.

Turn a pile of servers into one catalog you control.

Point your agent at one endpoint and you're governed in minutes — no code changes, no proxy to run.

Get started

Questions? Email us