Call this server's tools from your browser. Nothing to install.
Requests go through mcptry because browsers can't call most MCP servers directly. Header values are sent to the server for that request only and are never stored or logged.
Tool list from our last check. Connect live to refresh it.
Get one intent: required context, decision factors, honest supply status and policy. Never asserts prices or vendors.
Intent id, e.g. buy-vps or book-flight
This tool isn't marked read-only, so it may change something on the server or in an account you connected.
Pick your client. Values in capitals, like YOUR_TOKEN, are yours to fill in.
Run in a terminal
claude mcp add --transport http intent-catalog https://agentshopping.org/mcp
Get one intent: required context, decision factors, honest supply status and policy. Never asserts prices or vendors.
| Argument | Type | Description |
|---|
| id* | string | Intent id, e.g. buy-vps or book-flight |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "string",
"description": "Intent id, e.g. buy-vps or book-flight"
}
},
"additionalProperties": false
}List 100 buyer-agent transaction intents with supply status. 99 of 100 have no verified supply partner. Ranking is editorial, not measured demand.
No arguments.
{
"type": "object",
"properties": {},
"additionalProperties": false
}Supply catalog with dated verified facts and an explicit list of unverified fields. Currently one partner.
No arguments.
{
"type": "object",
"properties": {},
"additionalProperties": false
}Deterministic, explainable match or rejection for a workload. No scores, no invented numbers.
| Argument | Type | Description |
|---|---|---|
| workload* | string |
{
"type": "object",
"required": [
"workload"
],
"properties": {
"workload": {
"enum": [
"personal_proxy",
"background_jobs",
"self_hosted_automation",
"production_service",
"gpu_inference",
"managed_app"
],
"type": "string"
}
},
"additionalProperties": false
}We read every tool's name, description and inputs the way the model will, looking for instructions hidden from you. A clean result means no warning signs, not a guarantee. Scanned Nov 8, 2026.
No readOnlyHint / destructiveHint annotations, so clients can't tell a read from a write without asking.
Fix: Add annotations to each tool, e.g. { readOnlyHint: true } for tools that only read.