Think inside your AI world.

PostHog through Unl

PostHog answers “what are the numbers?” brilliantly. It doesn’t answer “are we there, by the bar I set, and why did I set it there?” — because that bar lives in your head, not in the dashboard. That is the read Unl adds, and PostHog is the first tool wired to it today.

PostHog through Unl reads your product analytics — trends, funnels, retention, HogQL — inside the milestone criteria you have already ratified, so the answer comes back as a verdict with its reasoning, not three loose numbers. It is live through Unl today, and the lane is open: one box, paste anything.

What PostHog holds

PostHog’s MCP server is one of the richest in the directory. The read surface is deep and honest:

  • Trends, funnels and retention curves
  • HogQL queries over the event stream
  • Saved insights and dashboards
  • Feature flags and experiments with their results
  • Cohorts, session data and error tracking
  • LLM cost tracking

What the naked read gives you

Point any MCP client at PostHog and you get exactly that data back — accurate, well-shaped, and silent on what it means for you. It will tell you activation is 38%. It will not tell you that 38% is the number blocking your milestone, because it has never been told what your milestone is.

The frame judges the data it is given; it does not verify the source’s accuracy.

What changes when PostHog is read measured

Say you set milestone two as three things at once: 100 weekly active users, activation at or above 40%, and week-four retention at or above 25% — and you set that activation floor deliberately, because below it the paid tier doesn’t pencil out. That last part is the bit no dashboard holds.

Asked in plain language — “are we at milestone two?” — the measured read returns: not met. WAUs and W4 clear your bars; activation at 38% is the one below the line, and it is the blocking metric because the reason you set that 40% floor was paid-tier viability. Same PostHog data; a verdict instead of a chart. And there is a sharper beat, from our own logs rather than a scenario: on 9 July we ran a live PostHog read through Unl and the first query suggested an internal:true flag was missing — which would have meant a data breach in our own instrumentation. The ratified quality doctrine we read against forces a raw-payload check before any breach is declared, so the check ran: the property was there all along. The alarm was ours, and it was false, and the rule caught it before we acted on it. That was the second such catch in two days on real data — which is what makes it a property of reading against your own criteria, not an anecdote.

And back again

If you decide the activation floor should hold at 40% rather than slip, you ratify that call back into Unl in the same conversation — so the next read of PostHog is measured against the criterion you just confirmed, and the milestone tightens rather than drifts.

The answer comes back measured against what you already decided, and why.

PostHog is live through Unl today, and the lane is open to every tool that speaks MCP: one box, paste anything.

Read further

Questions people ask

How do I connect PostHog to Claude, ChatGPT or Cursor?

PostHog ships its own MCP server, so any MCP client can read it directly for the raw data. Unl reaches PostHog the same way — over MCP — which means it works wherever you already work: Claude, Claude Code, Claude Desktop, ChatGPT and Cursor each add Unl as a connector, and the per-surface steps are set out at unlimitless.ai/connect, checked against each vendor’s own current documentation. Add the connector once and the same PostHog data arrives measured against the criteria you have ratified — a verdict with the why, not just the chart. PostHog is live through Unl today, and the lane is open to any tool that speaks MCP.

Can AI check my PostHog numbers against my targets?

Yes — that is the point of reading PostHog through Unl. You ratify the target (say activation at or above 40%, and why), and a plain-language question returns whether the live numbers meet it, which metric is blocking, and the reasoning you attached when you set the bar.

What does PostHog through Unl give me that a PostHog dashboard doesn’t?

A dashboard answers “what are the numbers?”. It cannot answer “are we there, by the bar I set, and why did I set it there?” — because that bar and its reasoning live in your head, not in PostHog. Reading PostHog through Unl adds the criteria layer: you ratify the milestone once, with the reason attached, and a plain-language question comes back as a verdict — met or not met, which metric is blocking, and why that bar exists. Same PostHog data, different answer shape.

Can I just use PostHog’s own MCP server instead?

Yes, for the raw read — PostHog’s MCP server is one of the richest available, and any MCP client can point at it and get accurate, well-shaped data back. What it returns is silent on what the numbers mean for you: it will tell you activation is 38%, but not that 38% is the number blocking your milestone, because it has never been told what your milestone is. That is the layer Unl adds on top of the same read, not a replacement for it.

Does Unl change or store my PostHog data?

Not stored: what Unl reads is quarantined for the turn and never kept — no query text and no read content reaches our telemetry. Not changed by a read, either — reading PostHog only reads it. Unl can call a PostHog tool that writes, because the socket exposes a server’s full surface rather than a chosen subset; but every write carries your explicit in-turn gesture, verified at the choke point, and a tool that doesn’t declare itself read-only is treated as a write and gated by default. It judges the data it is given; it does not verify the source’s own accuracy.

What this is

Think inside your AI world — you stay in command

Unlimitless (Unl to friends) holds what you've settled, reads what your tools are showing, and catches what's changed out in the world — and hands your AI whatever bears on the work, the moment it's needed, without you asking. The right thing, in front of the model, unprompted, with you in command of the call. So you keep moving toward what you set out to build, on top of everything you've already decided.

It plugs into Claude, Claude Code, ChatGPT and Cursor as an MCP connector. Quick to connect, in a couple of steps.

Unlimitless is open now to invited Alpha. Apply for the Beta waitlist to come in ahead of the full launch:

Alpha is invite-only · free at launch.