Your Ad Data. Inside Your AI.

Cresva is the only marketing intelligence platform that works natively inside Claude Desktop and ChatGPT. 9 real-time tools. Zero dashboards. Just ask.

Claude Desktop
Claude Desktop
MCP Protocol
ChatGPT
ChatGPT
Custom GPT on GPT Store
cresva.ai
Web Dashboard
9
Live tools
5
Ad platforms
<2s
Response time
0
Dashboards needed
Try on ChatGPTBook a Demo

The problem

Dashboards were built for a different era

Every marketing team runs the same loop: open a dashboard, filter to the right view, export to a spreadsheet, build a pivot table, screenshot it, paste it into Slack, and write a paragraph explaining what it means. That loop takes 10 to 15 minutes per question. Multiply by the dozens of questions your team asks every day, and you start to see the problem.

The data is there. The tools to access it are not. Marketing analytics has been stuck in the "pull" paradigm: you go to the data. AI-native analytics flips this. The data comes to you, inside the tools you already use for thinking, planning, and communicating.

Cresva is the first marketing platform built for this shift. Instead of another dashboard, we built an intelligence layer that plugs directly into Claude Desktop (via Model Context Protocol) and ChatGPT (via Custom GPT). You ask a question in natural language. You get an answer with real data in under 2 seconds.

Think where you think

Your AI assistant already handles strategy, writing, and analysis. Now it handles ad data too. No tab switching. No context loss.

10 seconds, not 15 minutes

Every question that used to require opening a dashboard, filtering, and exporting now takes a single sentence in Claude or ChatGPT.

Chain questions, not tabs

Follow-up questions are instant. Ask about ROAS, then drill into the underperforming campaign, then get a budget recommendation. One conversation.

Architecture

One backend. Every AI.

Cresva connects to your ad accounts once. Then the same intelligence layer serves Claude, ChatGPT, and the web app through a unified API.

Claude Desktop
Claude Desktop
MCP Protocol (stdio)
ChatGPT
ChatGPT
Custom GPT (REST)
cresva.ai
Web App (Next.js)
Cresva Intelligence Layer
cresva.ai/api/mcp
Production
list_brands
get_metrics
forecast
detect_anomalies
compare
budget_pacing
budget_recommendation
benchmarks
ask
Connected data sources
Meta AdsMeta Ads
Google AdsGoogle Ads
TikTok AdsTikTok Ads
ShopifyShopify
GA4GA4

Connect once, query everywhere

Link your Meta, Google, TikTok, Shopify, and GA4 accounts once. Cresva unifies the data into a single normalized schema. Every AI interface queries the same source of truth.

Secure by design

API key authentication on every request. All data encrypted in transit via HTTPS. Read-only access only. No AI can modify your campaigns, budgets, or account settings.

Real-time, not cached

Every query hits your live ad platform data. No stale exports, no daily syncs, no waiting for ETL pipelines. The answer you get at 3pm reflects what happened at 2:59pm.

ClaudeClaude Desktop

Native inside Claude via MCP

The Model Context Protocol (MCP) is an open standard by Anthropic that lets Claude Desktop connect to external tools natively. Cresva's MCP server gives Claude direct access to 9 marketing intelligence tools, so it can pull live ad data, run forecasts, and detect anomalies within the same conversation.

How it works

1
You ask a question
"What is my ROAS this week across all platforms?"
2
Claude selects the right tool
Claude reads your question, identifies it needs ad metrics, and calls cresva.get_metrics with the right parameters.
3
Cresva returns live data
The MCP server queries your connected ad accounts in real time and returns structured data to Claude.
4
Claude interprets and responds
Claude formats the data, identifies the key insight (e.g. TikTok underperforming), and suggests next steps.
Claude can chain multiple Cresva tools in one answer
Runs locally via stdio. Data never leaves your machine
Works with Claude's deep reasoning, not just data retrieval
Auto-discovers all 9 tools. No manual configuration per tool
Same API key works for Claude Desktop and ChatGPT
Setup in 3 minutes
claude_desktop_config.json
{
  "mcpServers": {
    "cresva": {
      "command": "npx",
      "args": ["-y", "cresva-mcp-server"],
      "env": {
        "CRESVA_API_KEY": "your-api-key-here",
        "CRESVA_BRAND_ID": "your-brand-id-here"
      }
    }
  }
}

Install with one command: npx cresva-mcp-server

npm
Example conversation
Claude Desktop
You

How is my ROAS across all platforms this week? And flag anything unusual.

Using cresva.get_metrics
Using cresva.detect_anomalies
Claude

Here is your ROAS for the last 7 days across all platforms:

Meta Ads3.2x ROAS
Google Ads2.9x ROAS
TikTok Ads1.8x ROAS
Blended2.8x ROAS
Anomaly detected

TikTok CPA spiked 42% on Tuesday vs the 7-day average. This coincides with a new ad set launch. The new creative is spending but not converting at the expected rate.

Want me to compare the new TikTok ad set against your top performers, or check the budget pacing impact?

Example conversation
ChatGPT
Cresva AI
GPT Store

Am I on track for my monthly budget? And what should I do if I am not?

Called getBudgetPacing
Called getBudgetRecommendation

CK Brand is pacing at 67% through the month with 72% of days elapsed.

MTD SpendA$43,200
Monthly TargetA$60,000
Projected EOMA$55,800
Recommendation

Increase Meta daily budget by 18% for the remaining 8 days. Meta is your most efficient channel at 3.2x ROAS and has headroom to scale. Keep Google steady. Pause TikTok ad set #3 (CPA 42% above target) and reallocate to TikTok ad set #1.

Want me to break this down by campaign, or forecast what hitting the full A$60K would mean for revenue?

ChatGPTChatGPT Custom GPT

Live on the GPT Store

Cresva AI is a published Custom GPT on OpenAI's GPT Store. It connects to the same backend as the MCP server, using 9 API actions to query your ad data in real time. Anyone on your team with the link can start asking questions immediately.

How it works

1
Open the Cresva AI GPT
Search for "Cresva AI" on the GPT Store, or use the direct link shared by your team.
2
Ask any question
Type a question in natural language. "Am I on track?" "What is my best campaign?" "Why did CPA spike?"
3
ChatGPT calls your API
The GPT selects the right actions (budget pacing, metrics, anomalies) and calls your Cresva backend.
4
Get a formatted answer
ChatGPT returns the data with context, charts, and follow-up suggestions. Ask follow-ups to drill deeper.
Published on the GPT Store. Searchable by anyone
9 actions, same API as the MCP server
Compatible with GPT-4o, GPT-4.5, and future models
Requires ChatGPT Plus or Team subscription
Share one link to onboard your whole team in seconds

Tool Reference

9 tools. Every detail.

Every tool listed below works identically in Claude Desktop and ChatGPT. Same parameters, same data, same response format.

List Brands

list_brands

Returns all connected brands with their linked ad platforms, account IDs, active status, and last sync timestamp. This is usually the first tool Claude or ChatGPT calls to establish context before running any analytics query.

Parameters

No required parameters

Example query

“What brands do I have connected?”

Returns

Brand name, connected platforms (Meta, Google, TikTok), account IDs, sync status

Get Metrics

get_metrics

The core analytics tool. Queries spend, revenue, ROAS, CPA, CTR, impressions, clicks, and conversions with flexible date ranges, platform filters, and campaign-level breakdowns. Supports daily, weekly, and monthly granularity.

Parameters

brand_id, date_range, platforms (optional), metrics (optional), breakdown (optional)

Example query

“What is my ROAS by platform for the last 30 days?”

Returns

Metrics by platform, campaign, or ad set with totals and period-over-period changes

Forecast

forecast

ML-powered predictions for spend, revenue, conversions, and ROAS over 14 to 90 day horizons. Uses historical performance data, seasonality patterns, and trend analysis. Returns point estimates with confidence intervals.

Parameters

brand_id, metric, horizon_days, platform (optional)

Example query

“Forecast my revenue for the next 30 days”

Returns

Daily predictions with 80% and 95% confidence intervals, trend direction, seasonal adjustments

Detect Anomalies

detect_anomalies

Statistical anomaly detection using z-score analysis across all active campaigns. Flags unusual changes in spend, CPA, ROAS, CTR, and conversion rate. Each anomaly includes severity, direction, and a plain-English explanation of what changed.

Parameters

brand_id, lookback_days (optional), sensitivity (optional)

Example query

“Are there any anomalies in my campaigns this week?”

Returns

List of anomalies with severity (low/medium/high), affected metric, magnitude, and likely cause

Compare

compare

Side-by-side comparison of campaigns, ad sets, or platforms across any metric. Useful for A/B testing analysis, platform efficiency comparison, and identifying top and bottom performers.

Parameters

brand_id, entity_type (campaign/adset/platform), entity_ids, metrics

Example query

“Compare my top 3 campaigns by conversions this month”

Returns

Comparison table with all requested metrics, percentage differences, and ranking

Budget Pacing

budget_pacing

Tracks month-to-date spend against monthly budget targets. Calculates current pacing percentage, projected end-of-month spend based on current daily run rate, and flags whether you are on track, underspending, or overspending.

Parameters

brand_id, month (optional, defaults to current)

Example query

“Am I on track for my monthly budget?”

Returns

MTD spend, monthly target, pacing %, projected EOM spend, status (on track / under / over)

Budget AI

budget_recommendation

AI-powered budget allocation recommendations based on historical performance, channel efficiency scores, and diminishing returns modeling. Suggests how to redistribute spend across platforms and campaigns to maximize ROAS or minimize CPA.

Parameters

brand_id, objective (maximize_roas/minimize_cpa), total_budget (optional)

Example query

“How should I allocate my budget across platforms?”

Returns

Recommended allocation per platform/campaign with expected impact on ROAS/CPA

Benchmarks

benchmarks

Compares your brand performance against industry averages for key metrics. Helps identify whether your CPA, ROAS, CTR, and conversion rates are above or below peers in your vertical.

Parameters

brand_id, industry (optional, auto-detected)

Example query

“How does my performance compare to industry benchmarks?”

Returns

Your metrics vs industry averages with percentile ranking and areas for improvement

Ask Anything

ask

Natural language query engine that accepts any free-form question about your ad performance. Interprets the intent, selects the right underlying tools, and returns a comprehensive answer. This is the catch-all for questions that do not fit neatly into one tool.

Parameters

brand_id, question (free text)

Example query

“What was my best performing campaign last month and why?”

Returns

Comprehensive answer synthesized from multiple data sources with supporting data

Workflows

One question. Multiple tools.

Claude and ChatGPT can chain multiple Cresva tools in a single response. Here are workflows real teams use every day.

Morning performance check

Start your day with a full picture in one question.

You ask

“Give me a morning briefing: ROAS across all platforms, any anomalies, and budget pacing status.”

get_metricsdetect_anomaliesbudget_pacing
You get

Blended ROAS, per-platform breakdown, flagged anomalies, pacing status, and suggested actions.

Campaign optimization

Find underperformers and get reallocation advice in one conversation.

You ask

“Compare my top 5 Meta campaigns by CPA, then recommend how to reallocate budget.”

comparebudget_recommendation
You get

Ranked campaign comparison, efficiency scores, and a specific reallocation plan with expected impact.

Forecasting with context

Predict future performance and understand what drives it.

You ask

“Forecast my revenue for the next 30 days. How does that compare to last month?”

forecastget_metrics
You get

30-day revenue forecast with confidence intervals, compared to last month's actual, with trend analysis.

Client reporting prep

Pull everything you need for a client update in one go.

You ask

“Summarize this month: total spend, ROAS by platform, top 3 campaigns, any issues, and where we stand vs budget.”

get_metricscomparedetect_anomaliesbudget_pacing
You get

Complete monthly summary with key metrics, winner campaigns, flagged issues, and budget status.

Security

Read-only by design

Cresva can analyze your data but never modify your campaigns. Every security decision was made with the assumption that AI access should be strictly observational.

API key authentication

Every request requires a valid API key. Keys are scoped to your account and can be rotated at any time from the Cresva dashboard.

Encrypted in transit

All communication between Claude, ChatGPT, and Cresva uses HTTPS/TLS encryption. No data is transmitted in plaintext.

Read-only access

AI integrations can query and analyze your ad data but cannot create, modify, pause, or delete any campaigns, ad sets, budgets, or account settings.

MCP runs locally

The Claude Desktop MCP server runs on your machine via stdio. Your data flows directly between Claude and Cresva without passing through any third-party infrastructure.

No data stored by AI

Neither Claude nor ChatGPT stores your ad data between sessions. Each query fetches fresh data from your Cresva account. When the conversation ends, the data is gone.

Team access control

Share the Custom GPT link with your team. Each team member authenticates with the same API key. Revoke access instantly by rotating the key.

Use cases

Built for people who ask questions all day

If your job involves asking “how are we doing?” more than once a day, this is for you.

Performance marketers

Check ROAS, CPA, and spend without opening a single dashboard. Get anomaly alerts and budget pacing in the same conversation you use for strategy.

“What is my CPA by campaign?”
“Any anomalies today?”
“Compare Meta vs Google ROAS”

Agency account managers

Pull client reporting data in seconds. Prep for status calls with one question instead of 30 minutes of spreadsheet work.

“Monthly summary for CK Brand”
“Are we pacing on budget?”
“Top 3 campaigns by revenue”

Marketing directors

Get the high-level view without waiting for someone to build a deck. Ask strategic questions and get answers backed by real-time data.

“How are we tracking vs forecast?”
“Where should we shift budget?”
“What is our blended ROAS trend?”

Founders and CMOs

Stay informed without micromanaging. Ask one question in Claude before your morning coffee and know exactly where you stand.

“Morning briefing on all brands”
“How does our performance compare to industry?”
“30-day revenue forecast”

The shift

Dashboards are dead

The best interface for your data is the AI you already use every day.

Before

Dashboards and spreadsheets

1.

Open dashboard. Wait for it to load.

2.

Navigate to the right view. Apply filters.

3.

Export to spreadsheet. Build a pivot table.

4.

Realize you need a different breakdown. Repeat.

5.

Screenshot. Paste into Slack. Write context.

~15 min
Per question, every time

After

Ask Claude or ChatGPT

Open Claude or ChatGPT (already open).

Ask: "How is my ROAS this week across all platforms?"

Get the answer with live data in 2 seconds.

Follow up: "Why did TikTok CPA spike?"

Get actionable insight. Move on with your day.

~10 sec
Per question, with follow-ups

FAQ

Frequently asked questions

Stop switching tabs.
Start asking.

The only marketing platform native to Claude Desktop and ChatGPT. Set up in 3 minutes. Ask your first question in 10 seconds.

Try on ChatGPTMCP Setup GuideBook a Demo
EncryptedAPI Key AuthRead-Only