Query it from inside your agent
Standing runs as an MCP server, so Claude Code, Claude Desktop, Cursor and anything else that speaks MCP can ask about your visibility directly. Same data, same plan limits, no dashboard required.
- Tools
- 4
- Transport
- HTTP
- To install
- Nothing
Connect it
A remote server, so there is nothing to run locally. Create a key on your account page — they begin standing_sk_ and are shown once.
claude mcp add --transport http standing https://usestanding.com/api/mcp \
--header "Authorization: Bearer standing_sk_..."{
"mcpServers": {
"standing": {
"type": "http",
"url": "https://usestanding.com/api/mcp",
"headers": {
"Authorization": "Bearer standing_sk_..."
}
}
}
}The four tools
Shaped for a model rather than a screen: flat JSON, explicit units, no HTML, and every number that carries uncertainty ships with its band.
check_visibilityScore, range and per-engine breakdown for any domainFreecompare_competitorsThe brands-by-engine grid for a categoryFreeget_visibility_historyThe trend, and what caused the last moveTrackget_fix_roadmapRanked citation gaps to closeOptimize
Why the range is part of the answer
This is the one thing worth knowing before wiring it into an agent.
These engines sample. Ask the same buying question five times and you get five slightly different answers, so a score that moved from 42 to 38 has usually not moved at all. Every tool here returns the interval alongside the number and a flag saying whether the move clears it.
That matters more for an agent than for a person. A human glancing at a bare number will often hesitate before calling it a trend. An agent will not — it will report the drop, and then recommend a fix for something that never happened.
Questions before you wire it in
No. It is a remote server over streamable HTTP, so there is nothing to run locally and nothing to keep updated. Point your client at the URL with an API key in the Authorization header.
Create an account and open the account page. Keys begin standing_sk_ and are shown once, so store it when it is issued. The free tier needs no card.
No, deliberately. The same entitlements apply as in the dashboard, because the service is the paywall rather than the interface. Calling a gated tool returns a normal tool result explaining what upgrading would buy, not a protocol error, so your agent can relay it rather than swallow it.
Because these engines sample: asking the same question five times does not give the same answer five times. A tool that returns score 38 alone will have an agent confidently report a 4-point drop as a result. Returning score 38 with range 33 to 48 and moveIsSignificant false stops that, which matters more for an agent than for a human, because an agent will not squint at the number and hesitate.
It can, and that costs us real money per call, so it is rate limited per key and per domain with a cooldown. Answers are cached by question within a week, which is why a second call about the same domain returns quickly.
Anything that speaks MCP over streamable HTTP with a bearer header. Claude Code and Claude Desktop are what we test against; Cursor and other clients use the same JSON configuration.
What this does not do
It reads measurements; it does not publish anything or change your site. Fixes are ranked and returned for you to apply, and no tool writes to any surface you own.