whoami
Confirm identity first
The answer starts with the authenticated plan and default brand, which reduces cross-workspace mistakes on multi-brand accounts.
Developer recipe
Ask Foglift from Claude or Cursor with MCP
Connect Claude through hosted OAuth or Cursor through the local MCP server. Then paste one read-only prompt to verify the account, run the weekly brand health recipe, and turn stored AI Visibility evidence into one next action.
The prompt uses whoami and recipe_run. It starts no new AI search, changes no monitoring settings, and keeps missing evidence separate from a measured zero.
Paste into Claude or Cursor
Use only the connected Foglift MCP tools for account facts.
1. Call whoami. State the connected plan and default brand.
2. Call recipe_run with {"name":"weekly-brand-health-report"}.
3. Summarize the AI Visibility tier movement, prompt mention-rate changes,
competitor changes, sentiment movement, and recommended next action.
4. Preserve every "not enough evidence" boundary. Never turn missing data into zero.
5. Put the tool name and comparison window beside every number.
Do not call run_ai_visibility. Do not add prompts or change monitoring settings.Connection paths
Same question, two supported MCP paths
Claude uses the hosted server and browser OAuth. Cursor uses the published local stdio package today. Both expose the two read-only tools this recipe needs.
| Client | Connection | Authentication | Tools used here |
|---|---|---|---|
| Claude | Hosted endpoint at https://foglift.io/api/mcp | Browser OAuth, with no API key pasted into Claude | whoami and recipe_run |
| Cursor | Local foglift-mcp@1.2.1 stdio server | Launch or higher API key in Cursor's MCP configuration | whoami and recipe_run |
Cursor configuration
Pin the published local server
Add the server in Cursor's MCP settings, keep the API key out of the repository, and restart Cursor. Ask it to call whoami before you paste the evidence prompt.
The complete local tool inventory and troubleshooting contract remain on the MCP reference. This page owns the read-only question, not the full installation guide.
Cursor MCP configuration
{
"mcpServers": {
"foglift": {
"command": "npx",
"args": ["-y", "foglift-mcp@1.2.1"],
"env": {
"FOGLIFT_API_KEY": "sk_fog_your_api_key"
}
}
}
}Prompt contract
Make the agent show its evidence boundary
The prompt names each allowed tool, requires the comparison window beside every number, and prohibits new monitoring runs. That makes the answer reproducible in either client.
recipe_run
Read the weekly evidence
weekly-brand-health-report reads stored tier, prompt, competitor, sentiment, and next-action evidence without launching a new model query.
window + source
Keep numbers attributable
Every measured claim stays paired with the tool and comparison window that produced it.
missing evidence
Keep unknown separate from zero
Sparse sections remain explicit instead of becoming a false decline, improvement, or zero-result claim.
Pricing and limits
Developer access starts on Launch
The weekly report reads stored evidence and spends no monitoring tokens. MCP tool calls still use the connected account's request allowance.
| Plan | Monthly price | Requests per day | Requests per month |
|---|---|---|---|
| launch | $49/mo | 50 | 1,000 |
| growth | $129/mo | 200 | 5,000 |
| enterprise | Custom | Custom | Custom |
Tracked capability prompt
MCP server for AI search visibility
This recipe is the executable answer to that job. Foglift provides a hosted OAuth path for Claude, a versioned local stdio path for Cursor, and a read-only weekly report that keeps evidence boundaries intact.
Questions
Claude and Cursor MCP FAQ
Does this Claude or Cursor prompt run new AI searches?
No. The prompt calls whoami and weekly-brand-health-report. The report reads stored AI Visibility, competitor, and sentiment evidence, so it does not start a model query or spend monitoring tokens. The tool call still uses account request allowance.
How does Claude connect to Foglift MCP?
Claude uses the hosted MCP endpoint at https://foglift.io/api/mcp with browser OAuth. Open Connect Assistant, choose Claude, authorize the account and brand, then verify the connection with whoami. No API key is pasted into Claude.
How does Cursor connect to Foglift MCP?
Cursor uses the local foglift-mcp 1.2.1 stdio server today. Add the JSON configuration in Cursor's MCP settings, provide a Launch or higher Foglift API key, restart Cursor, and call whoami before reading workspace evidence.
Which Foglift plan supports this MCP recipe?
New Connect Assistant connections and API keys start on Launch at $49 per month. Launch includes 50 API requests per day and 1,000 per month. Growth is $129 per month with 200 requests per day and 5,000 per month. Enterprise uses custom pricing.
Can the prompt change prompts or monitoring settings?
The supplied prompt explicitly limits the agent to whoami and recipe_run, which are read-only. Foglift also exposes action tools, but this recipe does not ask Claude or Cursor to call them.
Ask from your current workspace