PostHog + AI: Use PostHog Through Juma's Marketing Workspace
Connect PostHog to Juma and run conversion funnels, web-analytics briefs, retention analysis, and A/B-test management through chat. Finished assets, not chat responses. Strategy, taste, and judgment stay human.
Key takeaways
Yes, marketing teams can access PostHog through AI
Juma connects to PostHog through PostHog's official MCP server (open-source, posthog/mcp) and OAuth, so each marketer can query a project in plain English using their existing PostHog access.
Run the analytics-heavy PostHog jobs in chat
Conversion-funnel diagnosis, web-analytics traffic briefs, retention analysis by acquisition source, content and path performance, plus A/B-test and feature-flag management. Funnels and retention run on HogQL and insight-create underneath. Outputs ship as branded PDFs, Excel trackers, and ready-to-use copy briefs.
PostHog AI (the in-tool assistant formerly known as Max) stays inside PostHog
Juma works across the marketing toolchain: PostHog + HubSpot + Webflow + Google Ads in one chat, with the answer delivered as the deliverable.
400+ marketing teams use Juma
4.9/5 on G2. House of Growth ships 160 articles a month with 85+ hours saved.
Two pairings carry most marketing workflows on PostHog
PostHog + HubSpot for revenue attribution; PostHog + Webflow for the insight-to-action loop on landing pages.
Enterprise-grade by default
SOC 2 Type II (A-LIGN), ISO/IEC 27001, GDPR, HIPAA. AES-256 encryption. Zero data retention from AI providers.
Can I access PostHog through AI?
posthog/mcp on GitHub) and OAuth, so marketing teams can query PostHog projects in plain English and get back finished reports, not just chat answers. The connection runs at the user level, so each marketer's existing PostHog permissions carry directly into Juma without extra setup.The mechanism matters because PostHog's MCP server is the marketer-friendly path into a tool that has historically been engineer-leaning. PostHog itself acknowledged in early 2026 that PMs and marketers were using its in-tool AI assistant more than the ICP engineers it was built for. The MCP server opened up the same access from outside PostHog. Juma uses that server (plus OAuth where MCP isn't yet available) to build insights, run saved queries, and execute HogQL over events and persons on demand. Nothing is copied into a third-party warehouse, nothing is cached past the session, and every pull is logged in Juma's audit trail.
Once connected, PostHog works across every Juma project and chat, so an agency working across multiple clients doesn't need to remap per Juma Project. Trust signals live at the platform layer (zero data retention, AES-256, audit logs, SOC 2 Type II, ISO 27001, GDPR, HIPAA), unlimited seats, credit-based pricing, and the team reviews every output before it ships.
What can Juma do with PostHog?
Through Juma, marketing teams use PostHog for conversion-funnel diagnosis, web-analytics traffic briefs, retention analysis tied to acquisition source, content and path performance, and A/B-test and feature-flag management. Each one builds insights or runs HogQL through the connection and returns a finished deliverable: a branded PDF, an Excel tracker, or a ready-to-use copy brief, with human review built into each step.
01 · Conversion-funnel diagnosis
Conversion-funnel diagnosis
Describe a funnel in chat ("blog visit to signup to activation to paid trial") and Juma builds the funnel in PostHog with insight-create or HogQL, identifies the worst drop-off step, and writes a prioritized fix brief. PostHog's funnel-to-fix loop is the workflow its growth-marketing handbook leads with; Juma adds the synthesis layer that turns the numbers into a written brief the design and engineering team can act on. Strategy and taste stay with the team; assembly comes back in chat.
02 · Weekly web-analytics
Weekly web-analytics traffic brief
Run a weekly traffic brief from PostHog Web Analytics (the GA4 alternative PostHog launched in 2024). Compare paid, organic, direct, and referral side by side, with bounce rates, page-level conversion, and week-over-week deltas, computed through insights and HogQL. PostHog Web Analytics is first-party and not blocked by ad blockers, so the numbers are usually higher-fidelity than GA4's. Juma turns the raw numbers into a written summary with hypotheses for the team to validate.
03 · Retention analysis
Retention analysis by acquisition source
Ask in chat: "Do users we acquire from organic blog content retain better than users from paid search?" Juma builds the retention insight in PostHog and runs the HogQL behind it, then writes the answer with specific numbers and a recommendation for budget reallocation. Retention curves are split by acquisition source so the analysis stays current week to week without manual upkeep. AARRR-style reporting comes back as a branded PDF for the monthly review.
04 · Content and path
Content and path performance brief
Ask Juma to build PostHog's path-analysis and cross-domain funnel insights for the blog: which posts produce real signup intent, not just pageviews. Juma runs the paths and funnels through insight-create and HogQL, then writes a "what to refresh, what to retire, what to write next" brief based on the data. PostHog's own marketing team publishes content-performance numbers using PostHog. The team owns the editorial direction; the brief comes back drafted in chat.
05 · A/B-test and feature-flag
A/B-test and feature-flag management
Ask Juma to spin up an experiment ("test the new pricing-page headline against control") and it creates, launches, pauses, resumes, and concludes the PostHog experiment, then ships the winning variant once the team signs off. Juma also handles feature flags end to end: create, update, and check evaluation and status. PostHog runs experiments and flags in the same workspace as the analytics, so the test design, the rollout, and the readout all stay in one place. The call on what to ship stays human.
How does Juma + PostHog replace your product-analytics stack?
Marketing and growth teams typically stack PostHog with Mixpanel, Notion or Linear, and ChatGPT for ad-hoc synthesis on exported numbers. With Juma + PostHog, that middle layer disappears. PostHog stays the data source; what gets replaced is the toolchain between it and the deliverable, with funnels, retention, and paths computed through HogQL and insight-create. Strategy, taste, and judgment stay human.
Traditional product-analytics stack vs. Juma + PostHog
Why keep PostHog but drop the middle layer
Three real-world frustrations marketers post about publicly justify the layering. First, PostHog's documented critique is that it leans engineer-heavy: pulling the analysis still benefits from SQL or HogQL fluency. With Juma + PostHog, the team writes in chat and the SQL/HogQL happens underneath. Second, Mixpanel-vs-PostHog comparison content dominates the marketer search results in 2026 because most teams want consolidation but worry about losing features. Keeping PostHog and dropping the others is the cleanest consolidation play once Juma sits on top. Third, House of Growth ships 160 articles a month with 85+ hours saved monthly using Juma; The Crew sees 2x faster workflows with 90% adoption. The team didn't get smaller. The toolchain did.
What does PostHog work best with in Juma?
Marketing teams most often pair PostHog with HubSpot (for revenue attribution past the signup) and Webflow (for the insight-to-action loop on landing pages). Both pairings answer questions PostHog alone can't, and both run end-to-end inside one chat: connect each tool once, then ask in plain English for the analysis or the brief the team needs this week.
Revenue attribution
PostHog ends at the in-product event: signup, activation, feature use. It does not hold the lifecycle: MQL, SQL, closed-won, MRR, churn. HubSpot does. The pair closes the "did this campaign produce revenue, not just signups?" loop that B2B SaaS marketers ask every quarter. PostHog ships an official HubSpot destination and source, so the connection is stable and documented; the gap is the synthesis. Juma writes the source-by-source revenue scorecard from joined PostHog event data and HubSpot deal data, with a recommended budget shift the team's analyst reviews before sending.
The insight-to-action loop on landing pages
PostHog autocapture on a Webflow marketing site is a featured PostHog tutorial. Combined, the pair tells a marketer what's failing on a page (PostHog funnels and path analysis) and where to fix it (Webflow CMS). What's missing is the brief in between. Juma pulls the PostHog evidence, drafts the copy and design fix recommendations in the project's brand voice, and prepares the changes for the marketer to ship in Webflow. The handoff from insight to action stops being a Notion doc the team forgets about.
How to connect PostHog to Juma
Open Integrations
Click your name in the bottom-left corner of Juma, then Integrations.
Find PostHog
Under the Product Analytics section of the integration list.
Connect
Click Connect and either authorize via PostHog's OAuth or paste a Personal API key (find it at PostHog → Settings → Personal API keys). Juma uses PostHog's official MCP server for reads where available. The connection runs at the user level, so the PostHog projects you can access in PostHog become accessible through Juma.
Start using PostHog
PostHog is now available across every Juma project and chat. Juma will prompt to use PostHog when a Flow needs product-analytics data, or ask in chat directly: "Build last week's signup funnel from PostHog and write a fix brief for the biggest drop-off step."
Trusted by marketing teams that ship
Customer outcomes
240+ teams
Security & compliance
Questions, answered
Can I connect PostHog to ChatGPT or Claude?
Yes. PostHog ships an official open-source MCP server (posthog/mcp on GitHub) with Claude Code, Claude Desktop, and ChatGPT integration paths documented. Juma uses the same MCP server under the hood and adds the marketing layer: persistent project knowledge, brand voice, multi-tool orchestration, and finished deliverables. Teams that just want ad-hoc chat keep the raw MCP; teams that want repeatable weekly and monthly reports use Juma on top.
What can AI actually do with my PostHog data?
The five most common Juma jobs on PostHog: conversion-funnel diagnosis, weekly traffic briefs from Web Analytics, retention analysis by acquisition source, content and path performance briefs, and A/B-test and feature-flag management. Funnels and retention run on insight-create and HogQL underneath. Each comes back as a finished deliverable, not a chat answer. The team's analysts spend their time on strategy; the assembly is the part Juma takes off the plate.
Should I use PostHog or Mixpanel?
Both work for product analytics; PostHog bundles experiments, feature flags, and surveys into one workspace, while Mixpanel is more analytics-pure. Many teams pick PostHog when consolidation matters (fewer vendors, one data source) and Mixpanel when the analytics UX matters more than the bundle. Juma connects to either. The right answer depends on what else is in the team's stack.
Can Juma write my weekly product report from PostHog?
Yes. Once PostHog is connected, ask Juma in chat: "Write this week's product report covering signups by source, activation funnel performance, and retention deltas by acquisition source." Juma builds the insights and runs the HogQL behind them, then ships a branded PDF. The team reviews and adjusts before sending it to leadership or the client.
How is Juma different from PostHog AI (formerly Max)?
PostHog AI is the in-product assistant inside PostHog: useful for answering questions about a single PostHog project. Juma works across the marketing toolchain in one chat: PostHog + HubSpot + Webflow + Google Ads + Google Search Console, with the output delivered as the deliverable (PDF, Excel, copy brief). Use PostHog AI for fast in-PostHog questions; use Juma when the output is a deliverable the team or client will read.
Is PostHog hard to use for non-technical marketers?
PostHog has a documented learning-curve critique: pulling deeper analyses benefits from SQL or HogQL fluency, which most marketers don't have. PostHog AI helps inside the product; Juma helps from outside, across the toolchain. With Juma + PostHog, the marketer writes in plain English and the SQL/HogQL runs underneath. The team gets the analysis without the technical lift.
Can Juma run A/B tests and feature flags in PostHog?
Yes. Juma manages the full experiment lifecycle in PostHog: create, launch, pause, resume, and conclude an experiment, then ship the winning variant once the team approves. It also handles feature flags end to end (create, update, and check evaluation and status). PostHog runs experiments and flags in the same workspace as the analytics, so the test design, rollout, and readout stay in one place. The call on what to ship stays human.
Does Juma store my PostHog data?
No. Juma reads PostHog data through the official MCP server or API on demand and uses it within the session. Zero data retention from AI providers. AES-256 encryption in transit and at rest. All access is logged in Juma's audit trail. SOC 2 Type II, ISO/IEC 27001, GDPR, and HIPAA compliance posture sit underneath every read.
Ask AI about us
400+ marketing teams trust Juma. ChatGPT, Claude, and Perplexity know why. Ask them.
Ready when you are
What's on your plate today?
Start typing a task and Juma will figure out the rest.