Geodesy MCP — a quick guide

This is a short guide to connecting Claude directly to the GGSC data graph — the same underlying data behind the State of Geodesy 2026 assessment (capability maturity scores, tier risk analysis, the supply chain scenario models, and the full physical network inventory of observatories and stations). Once connected, you can ask questions about it directly in a normal conversation with Claude, and get answers pulled straight from the live dataset — no need to dig through the report tables by hand.

If you already have this connector set up: the address is changing. Please switch to the new URL below by Friday 2 October 2026 — the old connector will be decommissioned after that date. The simplest way: remove the old connector first, then add the new one below with its usual name. If you'd rather keep the old one working while you test the new one, give the new one a different name for now — e.g. "Geodesy MCP (new)" — so the two don't sit in your connector list looking identical with no way to tell which is which. You can rename it back to "Geodesy MCP" once you've removed the old one.

One-time setup

1. Go to claude.ai, open Settings → Connectors.
2. Click Add custom connector.
3. For the name, enter Geodesy MCP (or a temporarily distinct name — see the note above if you're keeping the old connector around for now). For the URL, enter:

   https://geodesy-mcp.merchantsofmalta.com/mcp

4. Leave everything under Advanced as-is and continue.
5. On the next screen ("Step 2 of 2"), you'll be asked two questions — the defaults Claude picks are already correct, so just confirm and move on:
   • Authentication: leave it on "Always required" (it should already be selected, marked Detected).
   • OAuth client: choose "No client ID — register one automatically" (also marked Detected). Don't pick "Use Anthropic's hosted client metadata" or enter a client ID manually — neither is needed here.
   • Additional request headers: leave blank.
6. Click through to finish adding the connector. Claude will then redirect you to sign in with GitHub — log in with your usual GitHub account, and approve the "Geodesy MCP" access request when GitHub asks.
7. You'll land back on Claude.ai with the connector showing as Connected.

That's it — you only need to do this once. After that, Claude remembers the connection for future conversations.

Using it

Just start a normal chat with Claude and ask a question. You don't need to know any special commands or syntax — Claude works out which tool to use on its own. For example, you could ask:

• "What's the current capability maturity score for Data Management?"
• "Which domain has the weakest average score, and why?"
• "Show me the tier risk scores for the Satellite Orbits supply chain."
• "Which capabilities are flagged as Phase 1 investment priorities?"
• "How many capabilities are assessed in the Governance and Strategy domain?"
• "Which institutions hold more than one infrastructure role right now?"
• "How many stations are there in the network, and how many of those are co-location sites?"
• "Which stations near [a region] have more than one measurement technique?"

Claude will query the live graph and answer using the actual current data — if something in the underlying assessment gets corrected or updated, your answer reflects that immediately, not a stale snapshot from when the report was published.

A few terms you'll see along the way

While Claude works, it narrates what it's doing — you may notice a few technical terms in that narration. You don't need to know what these mean to use it; they're just labels for what's happening behind the scenes:

• SPARQL query — the way Claude searches the data. Think of it like a search engine query, just built for this particular kind of structured data rather than web pages.
• Graph / dataset — another name for the database itself — the collection of GGSC information being searched.
• Triple — the basic unit the data is stored in. One triple is one fact — e.g. "Data Management → average score → 2.75." If Claude mentions a triple count, that's just a way of describing how much data it looked at.
• Station vs. co-location site — every physical tracking site in the network is a "station," whether it has one measurement instrument or several. A co-location site is specifically a station where two or more different techniques (e.g. GNSS and VLBI) sit at the same physical location — this is the industry's own term for it, not something specific to this project. So if you ask "how many stations are there," the answer includes co-location sites too; they're not a separate category sitting outside the count, just a more capable subset of it.

None of this changes how you use it — just ask your question in plain language, same as any other conversation with Claude.

What this connects to

This gives you access specifically to the UN-GGCE geodesy supply chain data — the same dataset behind the State of Geodesy 2026 assessment. It doesn't touch anything outside that scope. It covers seven areas (counts current as of 15 Sep 2026 — ask Claude for the latest any time, since these numbers can grow as the assessment is extended):

Data area	What's in it	Roughly how much
Capability maturity	The State of Geodesy 2026 scores themselves — People/Process/Technology/Data maturity plus an investment priority, for each capability	58 capabilities across 6 domains (Data Management 19, Governance and Strategy 14, Innovation and Development 9, Engagement and Collaboration 6, Asset and Infrastructure Management 5, Data Products and Software 5)
Physical network	Every observatory and station in the network — see the terms section below for what "co-location site" means	2,754 stations, 182 of them co-location sites
Infrastructure organisations	The institutions that actually run the network day to day — data centres, analysis centres, correlators, operation centres, technology-development centres, for VLBI and SLR so far	~120 organisational records. Eight institutions (e.g. GFZ Potsdam, NASA Goddard) turned up holding more than one of these roles at once across different techniques — flagged as a concentration-risk finding, not merged away. GNSS and DORIS don't have this organisational layer yet — a known, explicitly-flagged gap, not an oversight.
Tier risk and supply chain scenarios	Country-by-country risk exposure, plus the underlying workflow dependencies for Earth Orientation Parameters, Gravity Field determination, the Celestial and Terrestrial Reference Frames (ICRF/ITRF), and satellite orbit determination	~19 countries and regions modelled individually
Governance and organizations	Which bodies are involved and how they're related, plus the IAG Executive Committee roster	8 organisations, 25 named individuals. This one holds personal data (names and roles of real people) — it's governed by its own data-handling record, so if a question surfaces someone's name, that's expected, not a leak.
Event history	A running log of things that have actually happened (GNSS interference incidents, governance actions, orbit and gravity events) — this is what feeds the risk and maturity scores, not a static opinion	Smaller, growing over time

There's also a technical "rulebook" layer underneath all of this (the formal definitions and validation rules the data has to follow) — you'll never need to look at it directly; it's what keeps the areas above internally consistent.

If something doesn't work

If the connector won't connect, shows an error, or a question comes back with something that looks wrong or out of date, just let Ben know directly — happy to look into it.