Quickstart
LLMentions measures how AI answer engines (ChatGPT, Google AI Mode, Gemini, Claude, Perplexity and Grok) mention and cite your brand. It has three interfaces with the same capabilities: an MCP server for agents, a CLI, and a REST API.
1. Get a key
curl -s https://api.llmentions.com/v1/signup \
-H 'content-type: application/json' \
-d '{"name": "Acme"}'
You get a workspace and a test key (sk_test_…). Test keys run against mock engines with realistic variance and are never charged, so you can build and test your integration for free. A live key (sk_live_…) calls real engines and is charged per successful answer. See Billing.
The key is shown once. Keep it in your secrets manager:
export LLMENTIONS_API_KEY=sk_test_…
2. Ask a question
curl -s https://api.llmentions.com/v1/ai/answers \
-H "authorization: Bearer $LLMENTIONS_API_KEY" \
-H 'content-type: application/json' \
-d '{
"prompts": ["best seo api for ai agents"],
"engines": ["chatgpt", "perplexity", "claude"],
"samples": 5,
"brand": "Acme",
"domain": "acme.com",
"competitors": ["Rival"]
}'
You get back every answer in one normalized shape:
- the answer text
- citations, each classed as
owned,competitor,earned,ugcorreference - brand mentions, with the order of first appearance
- the search queries the engine ran
- where the answer came from (
api,uiorparametric)
Because you passed brand, you also get a report: mention rates with 95% confidence intervals.
One answer is one sample. AI answers vary from run to run, so use
samplesof 5 or more before you trust a rate. See Methodology.
Add "dry_run": true to get the price quote without running anything.
3. Connect your agent
claude mcp add --transport http llmentions https://api.llmentions.com/mcp \
--header "Authorization: Bearer $LLMENTIONS_API_KEY"
Then ask your agent: "Using llmentions, check how ChatGPT and Perplexity answer 'best seo api for ai agents' and whether Acme is mentioned." Setup for other clients is in MCP.
4. Track a brand over time
For repeated measurement, create a project with tracked prompts and run panels against it. Runs are stored and compared run-over-run with a significance test:
llmentions projects create Acme --domain acme.com --competitor Rival \
--prompt "best seo api for ai agents" --prompt "seo tools with an mcp server"
llmentions run Acme -n 5 --dry-run # quote
llmentions run Acme -n 5 # runs, waits, prints the report
llmentions answers Acme --missing # the answers where you were not mentioned
Where next
- MCP: clients and the tool catalog
- CLI: every command
- REST API: endpoints, errors and webhooks
- Billing: prices, holds and test keys
- Machine-readable:
/llms.txt,/llms-full.txt,/pricing.jsonand the OpenAPI spec