Connect Cursor or Claude to VectAd.
Automate the same audience workflow you run in the dashboard — describe segments in plain English, ground campaigns in your brand content, and activate on Meta, Google, and TikTok. No data science team required.
What your agent can do
Build audiences in plain English
Describe who you want — “customers like my best buyers” — preview match size, save the list, and sync to Meta, Google, or TikTok.
Ground campaigns in your brand
Pull in product pages and brand docs so media plans reflect your positioning, not generic templates.
Same engine as the dashboard
Connectors, customer data, and ad activation work identically whether you use the browser or your agent.
How it works
One loop from brand context to live audiences — the same path shown on the homepage.
Typical agent workflow
Connect Cursor or Claude and run the same audience workflow as the dashboard — from brand context to live ad platforms.
Add brand context
Pull product pages, brand guidelines, and key URLs into your workspace so plans reflect your actual positioning.
Spring collection page · positioning doc · category landing URL
Understand the opportunity
Ask plain-language questions over your content to surface who the offer is for and what claims you can make.
“Who is this product for?” · “What claims can we make?”
Ground the media plan
Generate a structured campaign — objective, channel mix, creatives, KPIs — from your own copy, not generic templates.
Plan grounded in your site and docs
Match real customers
Describe who you want in plain English, preview match size, and save a named audience from your connected data.
“High-value repeat buyers interested in premium outerwear”
Sync to ad platforms
Push the audience to Meta, Google, or TikTok. Only privacy-safe hashed identifiers leave your workspace.
Custom Audience live in Ads Manager · exclusion list on Google
The split: Brand context shapes the plan — positioning, creatives, channel strategy. Your customer data defines the list — who gets synced to Meta, Google, and TikTok.
- 01Check your workspace— Plan, synced profile counts, connected sources, and the single best next action.
- 02Connect your store— WooCommerce, Shopify, BigCommerce — paste REST credentials via MCP or use the web UI.
- 03Sync customers— Pull purchasers with LTV, recency, and product signals into your workspace.
- 04See what to do next— Data-driven recommendations (repeat buyers, lookalikes, win-back, etc.).
- 05Ingest brand context— Product pages, guidelines, competitor URLs → chunked corpus.
- 06Understand the opportunity— Ask “who is this for?” over your ingested pages.
- 07Ground the media plan— RAG-backed objective, channels, creatives, KPIs.
- 08Match real customers— Check which metrics apply (LTV vs engagement), describe your ideal segment in plain language, preview size, save the audience, and adjust precision vs reach for ad-platform minimums.
- 09Sync to ad platforms— Push hashed audience to Meta, Google, TikTok, etc.
Connect in Cursor
Connect in Cursor
VectAd works with Cursor over Streamable HTTP. Two options — API key is fastest; OAuth uses Cursor's built-in sign-in flow.
Option A — API key (recommended)
- Create a
vad_key in Settings (sign in → Settings → API keys). - Add the config below to
~/.cursor/mcp.jsonor your project.cursor/mcp.json(replace<paste-your-vad_-key-here>— replace the whole placeholder, do not append your key after it). - Open Cursor Settings → Tools & MCP and confirm vectad shows as connected. Tools appear under Available Tools in chat.
{
"mcpServers": {
"vectad": {
"url": "https://vectad.com/api/mcp",
"headers": {
"Authorization": "Bearer <paste-your-vad_-key-here>"
}
}
}
}Option B — OAuth in Cursor
- Add only the URL (no headers) to
~/.cursor/mcp.json:
{
"mcpServers": {
"vectad": {
"url": "https://vectad.com/api/mcp"
}
}
}- In Cursor Settings → Tools & MCP, find vectad and click Connect (or enable the server — Cursor opens your browser).
- Sign in to VectAd if prompted, then click Authorize.
- After you approve, the browser redirects to
localhostorcursor://…— that is expected. Close the browser tab and return to Cursor; the server should show as connected within a few seconds. - If it stays disconnected, open the Output panel (Ctrl+Shift+U / Cmd+Shift+U), choose MCP Logs, toggle the server off and on, or use Option A (API key) instead.
Claude.ai, ChatGPT, or Gemini web? Don't paste your key into OAuth fields.
Those web UIs only accept OAuth (or no-auth) for custom connectors — not a Bearer API key. Leave Client ID and Client Secret empty. MCP requires Pro ($249/mo) or higher — Starter plans cannot connect via OAuth or API key. Use Cursor, a CLI, or the mcp-remote bridge below once upgraded.
Claude.ai web Connectors are OAuth-only — leave Client ID and Secret empty. Use Cursor, Claude Code (CLI), or Claude Desktop (bridge below).
Web custom connectors accept only OAuth or No auth. Use a custom GPT with an API-Key (Bearer) Action, or the bridge below.
The consumer Gemini app has no custom-MCP UI. Use the Gemini CLI with an Authorization header.
CLI clients (Bearer header works directly)
Claude Code:
claude mcp add --transport http vectad https://vectad.com/api/mcp \
--header "Authorization: Bearer <paste-your-vad_-key-here>"Gemini CLI:
gemini mcp add --transport http vectad https://vectad.com/api/mcp \
-H "Authorization: Bearer <paste-your-vad_-key-here>"Universal bridge — for clients that take JSON config (Claude Desktop, others), mcp-remote attaches the Bearer header:
{
"mcpServers": {
"vectad": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://vectad.com/api/mcp",
"--header",
"Authorization: Bearer <paste-your-vad_-key-here>"
]
}
}
}Developer reference
Full tool catalog with parameters, live vs mock badges, and the developer workflow diagram — kept on a separate page so this guide stays outcome-first.
Open MCP API reference