AI setup
Set up Humblytics with Claude or Codex.
Paste one prompt. Your agent hands you one link to approve, sets up your account if you need one, and starts reading your funnel. You never copy a key.
Paste this into Claude Code or Codex
The prompt tells your agent how to connect itself. Works with or without an account.
## Humblytics, install
### 1. Connect the MCP server (start here)
You cannot complete a browser login, so connect over the programmatic flow.
Start authorization. No auth, no body:
POST https://app.humblytics.com/api/mcp/authorize
It returns { token, url }. Tell the user: "Open this URL to connect your
Humblytics account: <url>", then wait for them to confirm they have opened it.
Poll every 2 seconds until status is "approved". Give up after 5 minutes:
GET https://app.humblytics.com/api/mcp/poll/<token>
While pending, keep waiting and tell the user you are waiting for
authorization. The approved response includes { status: "approved", apiKey }.
Never print or log the apiKey.
Register the server with that key:
claude mcp add humblytics --transport http https://mcp.humblytics.com/v1 --header "Authorization: Bearer YOUR_API_KEY"
Replace YOUR_API_KEY with the apiKey from the poll step, then verify with
list_properties. Any MCP client works. Point it at https://mcp.humblytics.com/v1 over HTTP
transport, path /v1. The MCP is included on every plan.
No Humblytics account yet? Fetch the onboarding spec first and follow it to
create the account, add the site, and install the tracking script:
curl -o HUMBLYTICS.md https://app.humblytics.com/agent.md
### 2. What you get
Read access across the live funnel, and one narrow write surface.
Read: traffic summaries, trends and breakdowns, realtime, pages and page
details, entry and exit pages, clicks, forms, funnel queries, Sankey and funnel
suggestions, saved funnels, properties, split tests and their recommendations,
Meta Ads connections, campaigns, ads and daily insights, Google Ads connections
and campaigns. Revenue
attribution (get_ads_attribution) resolves once both an ad connector and a
revenue connector are connected.
Write, and only ever as a proposal the user approves: create_split_test,
update_split_test, stop_split_test, save_funnel, delete_saved_funnel. That is
the whole write surface; you cannot change anything else.
### 3. Rules
- Read before you write. Call get_split_test_recommendations before creating a test.
- Only ship a recommendation with confidence >= 70.
- Every write is a proposal. Show the change, then wait for an explicit yes
from the user before calling a write tool.
- Promoting a winning variant is the user's decision. Never promote one yourself.
- Default auto_stop_days: 14.
- goal: "click_through" for CTAs, "conversion" for forms.
### 4. Prefer raw HTTP? (secondary path)
The same surface is on a plain REST API. Use it if you are scripting outside an
MCP client.
Your API key is a secret. Do NOT paste it into CLAUDE.md, AGENTS.md, chat
messages, or any file that gets committed to git.
1. Create a `.env` file in your project root:
HUMBLYTICS_API_KEY=hmb_your_key_here
2. Add `.env` to `.gitignore` if it isn't already.
3. Load it into the shell your agent runs in (or use direnv / dotenv):
export $(grep -v '^#' .env | xargs)
4. Always reference the key as `$HUMBLYTICS_API_KEY`. Never inline the literal
value in a command you show the user or commit to the repo.
If `$HUMBLYTICS_API_KEY` is unset, stop and ask the user to run step 3. Do not
ask them to paste the key into chat.
API base: https://app.humblytics.com/api/external/v1
Auth header: Authorization: Bearer $HUMBLYTICS_API_KEY
Get recommendations:
GET /properties/{property_id}/split-test-recommendations?page={path}
Create a split test:
POST /properties/{property_id}/split-tests
Body: { name, page, type: "nocode", goal, auto_stop_days,
variants: [{ label, changes: [{ selector, op, value }] }] }
Example call (safe pattern):
curl -H "Authorization: Bearer $HUMBLYTICS_API_KEY" \
"https://app.humblytics.com/api/external/v1/properties/{property_id}/split-test-recommendations?page=/pricing"
The rules in section 3 apply on this path too.Already have an account?
On claude.ai or Claude Desktop, add the connector. In a terminal, run the command. Approve in the browser and you are connected.
claude mcp add humblytics --transport http https://mcp.humblytics.com/v1Any MCP client works. Point it at https://mcp.humblytics.com/v1 over streamable HTTP.
What your AI assistant can do.
Your assistant reads your real data. It cannot change anything until you approve it.
Reads
Answers questions from your live data.
- Traffic, sources and pages
- Funnels and where visitors leave
- Clicks and form submissions
- A/B test results
- Revenue by ad and source
Proposes, you approve
Changes wait for your OK before anything goes live.
- Create an A/B testNeeds approval
- Update or stop a testNeeds approval
- Save or remove a funnelNeeds approval
Applying a winning variant is always your decision.
Nothing goes live until you say so.
Your assistant drafts the test from your real data. Claude asks for your permission before it creates anything, and the test starts only when you allow it.
12 ready-made skills for your AI assistant.
5 use your live Humblytics data. 7 help with strategy and copy and need no connection. All are free to use and written as plain text files.
Read the repo on GitHubCRO Optimizer
MCPPulls live analytics through the Humblytics MCP, analyzes conversion funnels, identifies drop-off points, and generates prioritized A/B test hypotheses with ICE/RICE scoring.
- Live funnel analysis from Humblytics data
- Prioritized A/B test hypotheses with impact scores
- Drop-off detection with root cause analysis
- Copy and CTA optimization recommendations
A/B Test Generator
MCPReads your page and click analytics, generates test hypotheses with element selectors, and creates no-code split tests through the MCP once you approve them.
- Test hypothesis generation from live data
- Element selector targeting from click data
- Test creation via create_split_test, on your approval
- Auto-stop configuration and result monitoring
Heatmap Analyst
MCPAnalyzes click and element-level engagement data from Humblytics to surface UX friction, ignored CTAs, and dead zones. Generates prioritized optimization recommendations from what visitors actually click.
- Click heatmap + ignored-element detection
- Entry and exit page audits
- Dead-zone surfacing from click clusters
- Form field drop-off analysis
Revenue Attributor
MCPPairs Humblytics multi-touch attribution and Stripe revenue with Meta and Google Ads spend read through managed, read-only connectors, so there is no Meta App Review or Google Ads API token to arrange. Produces campaign ROAS breakdowns and budget reallocation plans.
- Campaign-to-revenue attribution across channels
- ROAS ranking with reallocation suggestions
- Organic vs paid vs direct performance comparison
- Spend analysis from connected Meta data
Funnel Reporter
MCPEnd-to-end SaaS funnel reporting from Humblytics analytics. Traffic, signups, trials, conversions, and revenue with period-over-period comparisons.
- Automated weekly and monthly funnel snapshots
- Traffic to signup to trial to paid tracking
- SaaS benchmark comparisons
- Board-ready report generation
Page CRO
Audit any landing page for conversion issues. Runs a 10-point framework covering first impression, value proposition, CTAs, social proof, and mobile experience.
- 10-point conversion audit framework
- Above-the-fold and CTA analysis
- Trust signal and social proof review
- Mobile experience and performance check
Email Sequences
Designs multi-email drip campaigns (onboarding, lead nurture, re-engagement, abandoned cart) with timing, branching logic, subject lines, and full copy. Applies Kennedy and Hormozi frameworks.
- Onboarding, nurture, and re-engagement flows
- Subject line and preview text optimization
- Branching logic from engagement triggers
- A/B test ideas for subject lines and CTAs
SEO Strategist
Keyword research, content gap analysis, on-page optimization, and technical SEO audits. Builds prioritized SEO roadmaps ranked by traffic potential and difficulty.
- Keyword research with intent clustering
- Content gap analysis vs competitors
- On-page and technical SEO audits
- Prioritized roadmap by traffic potential
Marketing Strategist
Full-stack marketing agent covering funnel design, go-to-market planning, conversion copywriting, and growth strategy. Hormozi, Schwartz, and Kennedy frameworks applied to your context.
- Funnel architecture and GTM planning
- Hormozi Value Equation pricing and offers
- Landing page strategy and positioning
- Growth loops and retention design
Copywriting
Conversion copywriting for landing pages, emails, and ads. Uses PAS, AIDA, and Before/After/Bridge frameworks to write copy that drives action.
- Landing page and hero section copy
- Email subject lines and sequences
- Ad headlines and descriptions
- CTA copy and microcopy
Ad Expert
Paid advertising across Meta, Google, TikTok, LinkedIn, and YouTube. Creative concepting, copywriting, audience targeting, and budget allocation. Pairs with Meta's own ads CLI to pause underperforming creatives, paused by default, so you approve before anything goes live.
- Multi-platform ad creative and copy
- Audience targeting and lookalike strategy
- Budget allocation across channels
- Pause / duplicate ads via Meta's `meta ads` CLI
Content Strategist
Multi-format content strategy covering articles, newsletters, social media, video scripts, and SEO. Audience psychology meets platform mechanics.
- SEO-driven content calendars
- Multi-format creation (blog, social, video)
- Content pyramid and repurposing workflows
- Performance analysis and optimization
Tasks your assistant can help with.
Use the same data without switching between tools.
Let your agent find and draft tests
Walk your high-traffic pages, pull recommendations, and get a test drafted on each one for you to approve, without opening a dashboard.
Post-deploy testing
Ship a new landing page, then ask your agent to propose a test against it while the context is still open.
Conversion recovery
When your agent spots a conversion drop, it pulls recommendations and drafts a recovery test for you to sign off on.
Bulk optimization
Read every client property in one workspace and queue up test proposals across all of them. Agencies, this is for you.
Use Humblytics with your other tools.
Your agent composes the Humblytics server with whatever else is on the PATH: Meta's meta ads CLI, the GitHub CLI, your CMS, your deploy tooling.
Connect Meta or Google Ads once at Connectors in the dashboard (read-only). The agent reads list_meta_campaigns, get_meta_daily_insights and get_ads_attribution, which resolves full-funnel ROAS once an ad connector and a revenue connector are both live. To manage Meta campaigns, it shells out to Meta's CLI.
Never give an agent direct Marketing API access through a system user on an unapproved developer app. Meta is banning ad accounts for it. Connectors or Meta's CLI only.
Read: Meta's Ads CLI for agents# 1. Read funnel (Humblytics MCP)
> get_traffic_summary
start=2026-04-01 end=2026-04-30
timezone=America/New_York
# 2. Draft a variant, you approve it
> create_split_test
# 3. Pull creative performance (Meta CLI)
$ meta ads insights get --campaign_id $CID \
--fields ad_id,ctr --output json
# 4. Pause the laggards (Meta CLI)
$ meta ads ad update $AD_ID \
--status PAUSED --no-inputTry Humblytics free for 14 days.
Add the script to your site, connect Stripe and see the revenue from each page. Then test your first change.
14-day trial · Card required to start · Cancel anytime
+59.4%