MCP server
Our comparison data in your AI workflow
Connect your AI assistant to our MCP server and ask about software for the site, installation or fleet. It gets scores, criteria and source links, and cites the page where every fact is stated. Free, no account.
https://www.fieldtested.eu/api/mcp/ Why
An assistant that gets facts with a source makes up less.
Without a source a language model guesses at prices, features and local rules. With our server it gets the published score, every criterion with its source link and check date, and "unknown" where we know nothing.
What it looks like
Real answers from our tools on today's data. Your assistant turns them into an answer with links.
Find software for work orders.
Tool: search({ query: "work order", lang: "en" })
[
{
"type": "category",
"id": "work-orders",
"name": "Work order apps for installers",
"cluster": "Construction and installation",
"tools": 9,
"url": "https://www.fieldtested.eu/en/work-order-app/"
}
] Which are the highest scoring tools in this category?
Tool: get_category_ranking({ category: "work-orders" })
{
"category": {
"id": "work-orders",
"name": "Work order apps for installers",
"url": "https://www.fieldtested.eu/en/work-order-app/",
"rankingUrl": "https://www.fieldtested.eu/en/work-order-app/ranking/"
},
"ranking": [
{
"rank": 1,
"name": "Dynamics 365 Field Service",
"score": 100,
"coverage": 100,
"insufficientData": false,
"url": "https://www.fieldtested.eu/en/software/dynamics-365-field-service/"
},
{
"rank": 2,
"name": "Robaws",
"score": 100,
"coverage": 100,
"insufficientData": false,
"url": "https://www.fieldtested.eu/en/software/robaws/"
},
{
"rank": 3,
"name": "Cafca Software",
"score": 92,
"coverage": 92,
"insufficientData": false,
"url": "https://www.fieldtested.eu/en/software/cafca-software/"
}
]
} Compare the first two.
Tool: compare_tools({ ids: ["dynamics-365-field-service", "robaws"] })
{
"scores": [
{
"id": "dynamics-365-field-service",
"score": 100,
"rank": 1
},
{
"id": "robaws",
"score": 100,
"rank": 2
}
],
"criteria": [
{
"id": "digital_work_order",
"label": "Digital work order with customer signature",
"weight": 3,
"values": [
{
"id": "dynamics-365-field-service",
"value": "yes",
"evidence": "https://learn.microsoft.com/en-us/dynamics365/field-service/mobile/create-service-report",
"checkedAt": "unknown"
},
{
"id": "robaws",
"value": "yes",
"evidence": "https://robaws.com/nl-BE/functionaliteiten/digitale-werkbon",
"checkedAt": "unknown"
}
]
},
{
"id": "scheduling",
"label": "Technician scheduling in the same tool",
"weight": 2,
"values": [
{
"id": "dynamics-365-field-service",
"value": "yes",
"evidence": "https://www.microsoft.com/nl-be/dynamics-365/products/field-service",
"checkedAt": "unknown"
},
{
"id": "robaws",
"value": "yes",
"evidence": "https://robaws.com/nl-BE/functionaliteiten/digitale-werkbon",
"checkedAt": "unknown"
}
]
}
],
"compareUrl": "https://www.fieldtested.eu/en/compare/dynamics-365-field-service-vs-robaws/"
} Five tools, read only
The server writes nothing and keeps no questions. Every answer contains the URL to cite.
-
search
Searches categories and tools by name, in English, Dutch and French.
-
get_category_ranking
The ranking of a category: score, position, share with a source, label when data is insufficient, criteria and weights.
-
get_tool
The profile of a tool: price, languages, and per category every criterion with value, source link and check date.
-
compare_tools
Two to four tools side by side in the categories they share.
-
get_method
How we score, what does not count and how to cite us.
Available in your AI stack
The server speaks the MCP protocol over Streamable HTTP (versions 2025-03-26 to 2026-07-28). Menu names differ per version of each app. Look for "connector" or "MCP server".
-
Claude (claude.ai and Claude Desktop)
Go to Settings, Connectors, and add a custom connector with the server URL.
https://www.fieldtested.eu/api/mcp/ -
Claude Code
Add the server from the terminal.
claude mcp add --transport http fieldtested https://www.fieldtested.eu/api/mcp/
-
ChatGPT
In ChatGPT's developer mode you can add your own connector with an MCP URL.
https://www.fieldtested.eu/api/mcp/ -
GitHub Copilot in VS Code
Put this in .vscode/mcp.json.
{ "servers": { "fieldtested": { "type": "http", "url": "https://www.fieldtested.eu/api/mcp/" } } } -
Cursor and other clients
Most clients read an mcpServers block. Clients that only know local servers use mcp-remote as a bridge.
{ "mcpServers": { "fieldtested": { "url": "https://www.fieldtested.eu/api/mcp/" } } } -
Directly
Without an AI client: send JSON-RPC with a POST. The same data as the open dataset.
curl -s https://www.fieldtested.eu/api/mcp/ \ -H "Content-Type: application/json" \ -H "Accept: application/json, text/event-stream" \ -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"search","arguments":{"query":"work order"}}}'
Buyer signals via MCP Coming soon
Vendors with a Signals plan will later also get their buyer signals via MCP, straight into Claude, ChatGPT or their CRM assistant. That requires signing in through the vendor portal and does not exist yet. Leave your details and you hear about it first.
Frequently asked questions
What does it cost?
Nothing. The server is public and read only. The data is under CC BY 4.0: name the source and link to the page.
Are there limits?
Yes, 60 calls per minute per IP address. Need more? Use the open dataset in /data/v1/ or e-mail us.
Does the server contain personal data?
No. Only published data about software: scores, criteria, prices and sources. There are no buyer signals and no data about visitors.
Can a vendor rank higher through MCP?
No. The server shows the same score and order as the site, from the same calculation. Paying never changes them.
How up to date is the data?
As up to date as the site: every update of the site updates the server too. Every data point has a check date.
Access to signals via MCP
Tell us for which categories and in which AI tool you want to use signals. We contact you as soon as it is possible.