GenSig for agents: your AI visibility numbers over MCP

Claude

Claude Code Cursor Your agent

GenSig

ChatGPT4/12

Gemini6/12

Perplexity2/12

Example. One report per assistant, with how many of your questions named you.

One URL. Nothing to install.

No terminal needed in Claude: add it as a connector and sign in. Anywhere else, paste an API key.

In your agents

# Settings → Connectors → Add custom connector
Name   GenSig
URL    https://app.gensig.app/mcp
# Sign in to GenSig and approve. Done.

$ claude mcp add --transport http gensig https://app.gensig.app/mcp \
    --header "Authorization: Bearer gsk_live_…"
# then ask: "How did my brand do in Gemini this month?"

// ~/.cursor/mcp.json
{
  "mcpServers": {
    "gensig": {
      "url": "https://app.gensig.app/mcp",
      "headers": { "Authorization": "Bearer gsk_live_…" }
    }
  }
}

# read your reports
$ curl https://app.gensig.app/api/reports \
    -H "Authorization: Bearer gsk_live_…"
# start an analysis (key with the run scope, uses 1 analysis per assistant, 3 for Claude)
$ curl -X POST https://app.gensig.app/api/analyze \
    -H "Authorization: Bearer gsk_live_…" \
    -H "Content-Type: application/json" \
    -d '{"url": "cafetora.example", "surfaces": ["gemini"]}'

Reads your latest finished report, with its date. Create keys in the app under Settings; each key is shown once.

Nine tools. All read-only.

Your agent picks the one it needs.

get_report

The latest report, brand by brand and assistant by assistant.

{ "brand": "Cafetora", "named": 4, "answers": 12,
  "ci95": [13.8, 60.9], "surfaces": { "chatgpt": … },
  "measured_at": "2026-11-03" }

list_brands

Every brand in your workspace, with its last result.

→ domain · named / answers · ci95 · last_measured_at

get_visibility_history

One point per report, newest first.

→ measured_at · surfaces · named / answers · ci95

get_questions

Your questions, and who got named in each one.

→ question · target_mentions · leader

get_competitors

Who AI names in your category, sorted.

→ target · competitors[] · ci95

get_content_opportunities

Questions where you're missing or behind.

→ question · winning_brand · priority

get_action_plan

Where to act first: weakest intents, content gaps and the sources AI cites.

→ weakest_intents · top_content_gaps · cited_source_domains

list_content_briefs

The questions that have a brief ready to write from.

→ question · at

get_content_brief

The brief for one question: who AI names instead, the sites it reads, the shape.

→ question · brief

Run scopePOST /api/analyze

Start a new measurement from a script. Uses analyses like the app.

Your agent gets the margin of error too.

So it can say "likely 2 to 7 of 12" instead of guessing a trend.

AI rarely answers the same way twice. We count answers and show how sure each number is.

Example

"brand": "Cafetora",
"named": 4, "answers": 12,
"ci95": [13.8, 60.9]

4of 12 answers · ChatGPT

Ideas to wire it into.

Your agent, your tools. GenSig brings the numbers.

Monday report

"Every Monday, post how each assistant named us last week, with the range."

list_brands→get_visibility_history

Content brief

"Draft a brief for the question where Roastline beats us most."

get_content_opportunities→get_questions

Every client, one key

"Which of my clients lost ground this month?"

A key reads every brand in your workspace. Don't put it in a tool a client can talk to.

list_brands→get_competitors

> Which question do we lose most on Gemini?> Compare ChatGPT and Perplexity for us> Which sites does AI cite in our category?> Summarize this month for the client call> Did the new guide change anything yet?> Who leads in Claude, and by how much? > Which question do we lose most on Gemini?> Compare ChatGPT and Perplexity for us> Which sites does AI cite in our category?> Summarize this month for the client call> Did the new guide change anything yet?> Who leads in Claude, and by how much?

Read-only by default.

mcp: Reads finished reports. Never starts an analysis.
data: Your latest report and the date it was measured. Nothing is measured live.
auth: OAuth 2.1 with PKCE for the Claude connector.
keys: Stored as a hash, shown once, revocable any time.
scopes: read by default. run starts analyses and uses them like the app.
reach: A key reads every brand in your workspace. Keep it out of agents your clients talk to.
limits: 60 requests a minute per key. Keys can't create keys or delete data.
price: Included in every plan.

Give your agent the numbers.

MCPhttps://app.gensig.app/mcp

Create your account See an example report →

→ tools/list
list_brands
get_report
get_visibility_history
get_questions
get_competitors
get_content_opportunities
get_action_plan
list_content_briefs
get_content_brief
9 tools · read-only

Before you connect

Does my agent use my analyses?

Not through MCP. The MCP server only reads finished reports and never starts an analysis. Only an API key with the run scope can start one, and it uses analyses exactly like the app.

Which clients work?

Any client that speaks remote MCP over HTTP: the Claude connector (sign in with OAuth), Claude Code and Cursor (with an API key), and agents you build yourself.

What can the agent see?

Only what's in your finished reports, and only when you ask: the brands, reports, questions, competitors and action plans of the workspace that approved it. It passes through your agent's provider, like anything you paste into it. It cannot change settings, create keys or delete anything.

How fresh are the numbers?

The agent reads your latest finished report, and every answer carries the date it was measured. To refresh, run a new analysis in the app, let your monitor run, or start one with an API key that has the run scope.

Where does my data go?

What the tools return goes to your agent's provider, like anything you paste into a chat with it. If you work under an NDA, check that provider's data terms before connecting.

Does it cost extra?

No. MCP and API keys are included in every plan. A key with the run scope uses your analyses, like the app.