MCP Integration | Tailor AI

What is MCP?

MCP (Model Context Protocol) is an open standard that lets AI assistants connect to external tools and data sources. When you connect Tailor via MCP, your AI assistant can read your traffic data, review experiment results, query enrichment data, pull ads performance, and even take actions like starting or stopping experiments.

This means you can ask questions like "Which campaigns drove the most conversions last week?" or "Show me enterprise visitors who hit our pricing page" and get answers directly from your Tailor data, without switching tools.

Note: Tailor's MCP integration is a work in progress and likely to evolve as we add new capabilities. Claude's MCP configuration flows may also change over time. The setup details below are meant as examples only. If something looks different from what you see, check the Tailor settings page or reach out to support.

Setup

Step 1: Find your MCP URL

Go to Settings → Integrations. Your MCP URL is shown under Connected Apps, and an API key (for headless or scripted use) can be generated under API Access.

https://mcp.tailorhq.ai

Step 2: Connect your client

Claude CodeOAuth or API key▶

Run this in your terminal. You'll be prompted to authorize access in your browser the first time.

claude mcp add tailor https://mcp.tailorhq.ai --transport http

You can also click How to connect Claude Code in the Connected Apps panel for an in-app, copy-paste version of this command and the headless API-key variant.

Using an API key instead (for CI/CD or headless environments)

claude mcp add tailor https://mcp.tailorhq.ai --transport http --header "Authorization: Bearer YOUR_API_KEY"

Claude Desktop or claude.aiOAuth flow▶

This walkthrough uses the Claude desktop app and the OAuth flow. For claude.ai look for Connectors in your settings; for API-key auth, see the edit the config file directly option below.

On an enterprise Claude account? Your workspace admin may need to approve or add the Tailor connector before you can complete these steps. If the + button is missing or your connector doesn't appear after adding, reach out to your Claude admin or use the edit the config file directly option at the bottom of this section.

1. Open Customize → Connectors and click the + button in the top right.
2. Enter https://mcp.tailorhq.ai and click Connect.
3. Authorize Claude to access your Tailor AI account. Read and Write are toggled separately in Tailor; you can tighten them later in Settings → Integrations → Connected Apps.
4. Done. Ask Claude something like "show me the identified visitors from Tailor". The first time Claude calls a Tailor tool, it will ask for permission.

Alternative: edit the config file directly

Open Claude Desktop → Settings → Developer → Edit Config and add:

{
  "mcpServers": {
    "tailor": {
      "url": "https://mcp.tailorhq.ai",
      "transport": "http"
    }
  }
}

For API key auth, add "headers": { "Authorization": "Bearer YOUR_API_KEY" } inside the tailor object.

Other MCP clients▶

Any MCP-compatible client (Cursor, Windsurf, custom agents) can connect using https://mcp.tailorhq.ai with HTTP transport and either OAuth or an API key.

Step 3: Verify

Once connected, you'll see your connection with a CONNECTED status in Settings → Integrations → Connected Apps. Each connection has independent Read and Write toggles.

Available tools

All tools are available automatically once connected. Click a category to see what's included.

Traffic and analytics▶

tailor_traffic — Breakdowns by source, campaign, keyword, device, referrer, or locale with conversion rates
tailor_conversions — Conversion goal rates, most-clicked elements, and form submissions
tailor_engagement — Scroll depth and time-on-page metrics per landing page
tailor_visitor_journey — Full event timeline for a specific visitor

Experiments▶

tailor_experiments — List and view A/B test experiments
tailor_variants — Inspect individual variant configurations and copy changes
tailor_metrics — Per-variant performance (impressions, conversions, conversion rates)
tailor_experiment_downstream — Per-variant breakdowns of CTAs, goals, engagement, and audience demographics for a specific experiment
tailor_goals — List conversion goals configured for experiments
tailor_recommendations — AI-generated next-best-action recommendations

AI page tailoring▶

tailor_scrape_page — Inspect a landing page and return its tailorable text elements with stable signatures
tailor_infer_copy_changes — Generate audience-targeted copy proposals for a page (returns a token to apply via tailor_apply_proposals)

Visitor enrichment▶

tailor_enrichment — Company-level visitor data (industry, size, revenue, role, location)
tailor_usage — Enrichment credit usage and daily volume metrics

Ads and analytics platforms▶

tailor_google_ads — Google Ads campaigns, ad groups, ads, keywords, search terms, and landing page spend
tailor_meta_ads — Meta (Facebook/Instagram) campaigns, ad sets, and ads performance
tailor_amplitude — Amplitude event analytics and cross-platform experiment attribution

Monitoring and alerts▶

tailor_watchdog — Watchdog alerts and custom rules for ad spend and traffic anomalies
tailor_inbox — Pending action items (recommendations, alerts, experiment issues)
tailor_intent_mismatch — Pages where paid or organic search intent doesn't match the landing page headline, ranked by CVR deficit
tailor_competitors — Monitor competitor domains, URLs, and page changes
tailor_competitor_reports — AI-generated competitor analysis reports

Write actions (requires Write permission)▶

tailor_experiment_actions — Start, stop, or ramp experiments
tailor_experiment_setup — Create experiments, set primary goals, delete experiments
tailor_variant_manage — Create, delete, or restore variants
tailor_goal_manage — Create, update, delete, or attach conversion goals
tailor_recommendation_actions — Accept or dismiss AI recommendations
tailor_apply_proposals — Apply AI copy proposals (from tailor_infer_copy_changes) to a variant by token

Flexible query▶

tailor_query — Structured analytics queries with custom groupings, filters, and aggregations on Tailor or Amplitude data

Account and integrations▶

tailor_organization — Organization details, Slack settings, enrichment config
tailor_slack — Slack integration status and channel configuration
tailor_landing_pages — All landing pages being tested or optimized
tailor_dashboard — High-level summary: active experiments, pending recommendations, recent winners
tailor_suggested_questions — Context-aware starter prompts per dashboard surface (powers in-app chat suggestions)

Example prompts

Try these in Claude Code, Claude Desktop, or any MCP-connected assistant.

Review experiment results

"Review all active experiments. For each one, show me the variant performance, conversion rates, and whether Tailor has a recommendation to ramp or stop."

Tools: tailor_experiments, tailor_metrics, tailor_recommendations

Full-funnel data analysis

"Compare Google Ads and Meta Ads performance for the last 30 days. Show me spend, landing page traffic, and downstream Amplitude conversions. Which campaigns have the best cost per qualified conversion?"

Tools: tailor_google_ads, tailor_meta_ads, tailor_traffic, tailor_amplitude

Identify and qualify enterprise visitors

"Pull identified visitors from /enterprise and /contact-sales for the last 7 days. Filter to enterprise companies in software or financial services. Score by engagement and show company name, industry, size, and visitor role. Rank by fit."

Tools: tailor_enrichment, tailor_engagement, tailor_query

Monitor performance and catch issues

"Check if there are any Watchdog alerts or inbox items I should look at. If there are traffic anomalies, dig into which campaigns or sources are affected."

Tools: tailor_watchdog, tailor_inbox, tailor_traffic

Find pages where ad intent doesn't match the headline

"Show me pages where paid-search intent doesn't match the landing page headline. Which keyword themes have the biggest conversion-rate deficit, and what should I test on those pages?"

Tools: tailor_intent_mismatch, tailor_recommendations, tailor_traffic

Dig into how an experiment shifted behavior

"For our pricing-page-headline experiment, break down per-variant performance across all conversion goals, and show how audience demographics differ between variants."

Tools: tailor_experiment_downstream, tailor_metrics

Competitive intelligence

"Show me recent changes to competitor landing pages. What messaging or positioning shifts should I know about?"

Tools: tailor_competitors, tailor_competitor_reports

Building agentic workflows

Beyond interactive use, you can embed Tailor's MCP tools into automated pipelines using any MCP client library with an API key. Some ideas:

Weekly experiment digest

Pull active experiment results and pending recommendations, post a summary to Slack with suggested next actions.

Lead qualification pipeline

Pull identified visitors from high-intent pages daily, filter by ICP criteria, score by engagement, push qualified leads to CRM or Slack.

Cross-platform performance report

Combine Google Ads spend, Tailor site traffic, and Amplitude downstream conversions into a unified cost-per-conversion report.

Automated experiment management

Monitor results daily. When Tailor recommends ramping a winner, ramp it automatically with safety checks and Slack notifications.

Permissions and safety

Each connection has independent Read and Write toggles. Destructive write actions (stopping or deleting experiments) require a two-step confirmation token to prevent accidental changes.

Interactive analysis

Start with Read only. Enable Write when you're ready to take actions.

Automated reporting

Read only. No risk of accidental changes.

Experiment management

Read + Write, with confirmation steps and Slack notifications in your workflow.

Troubleshooting

"Tool not found" or no tools available

Verify the connection shows as CONNECTED in Settings → Integrations → Connected Apps. If not, re-run the connection command.

Write actions are rejected

Check that the Write toggle is enabled for your connection in Settings → Integrations → Connected Apps.

No data returned

Most tools require a date range. Make sure you're querying dates with active traffic, and that the relevant integration (Google Ads, Meta Ads, Amplitude) is connected.