Get a weekly readout of your running PostHog experiments

Every Monday: each running PostHog experiment's variants, exposures, conversion and chance to win, how long it has run, and which are ready to call.

Runs
Mondays at 9:00
Uses
PostHog
Posts
Report card

From your apps to your board

How the skill works: your agent reads running experiments from PostHog; posthog project id, posthog host from What you tell it, mondays at 9:00, and gives you report card and the report in the chat.

Reads

  • PostHogRunning experiments
  • What you tell itPostHog project ID, PostHog host
reads
Your agentClaude Code, Codex, Cursor or GeminiMondays at 9:00
posts

You get

  • Report cardOn your board, updated in place every run
  • The report in the chatEvery run, with or without Briefkin

What you get

  • Running experiments and how many have a result.
  • Per variant: exposures, conversion, chance to win or p-value, days running.
  • Flags: winners, experiments running long without a result, uneven exposure splits.
  • One line per experiment: ship, end, or keep running.

The card on the right is a sample. Yours shows your own numbers and updates in place every run.

What lands on your board

AcmePostHog experiment results · Claude Code

1 of 4 experiments has a winner

4 experiments are running. New pricing page has a 96% chance to beat control on checkout started. Onboarding checklist has run 31 days without a clear result.

ExperimentVariantExposuresConversionChance to winDays
New pricing pagetest4,2106.8%96%18
New pricing pagecontrol4,1855.9%4%18
Onboarding checklisttest2,94041.2%58%31
Onboarding checklistcontrol2,98840.6%42%31
Annual plan nudgetest1,1203.1%71%9
Annual plan nudgecontrol1,0982.6%29%9
Shorter signup formtest61022.0%64%3
Shorter signup formcontrol59821.1%36%3

Ship the new pricing page. End the onboarding checklist test: after 31 days the gap is 0.6 points and still undecided.

Done
An example with made-up numbers. Your agent posts your real ones.

What it does

Experiments get launched with energy and then forgotten. The one that reached a clear result two weeks ago keeps holding half your users on the old version, and the one that will never reach significance keeps running because nobody looked.

Every Monday the agent reads each running experiment's results from PostHog and posts one table: variants, exposures, conversion on the primary metric, the chance to win (or p-value), and days running. Product managers and engineers see which experiments to ship, which to end, and which need more time.

How it works

  1. 1List running experiments

    Call GET /api/projects/:project_id/experiments/ (or experiment-list on the PostHog MCP server) and keep the experiments with a start_date and no end_date, which is what PostHog shows as running. Note name, feature_flag_key, start_date and the primary metrics.

  2. 2Get the results

    On the PostHog MCP server, call experiment-results-get for each experiment (experiment-metrics-recalculation-latest-retrieve returns the last computed run without starting a new one). Over REST, start a run with POST /api/projects/:project_id/experiments/:id/metrics_recalculation/ and poll GET .../metrics_recalculation/:recalculation_id/ until its results are complete.

  3. 3

    3Read each experiment

    Per variant: exposures, conversion on the primary metric, and the statistic the experiment uses (chance to win for Bayesian, p-value and interval for frequentist). Flag a winner, an experiment over 28 days without a result, and exposures more than 10% away from the planned split.

  4. 4

    4Post the readout

    One row per variant, experiments with a decision first, and one line per experiment on what to do: ship, end, or keep running.

  5. Post it to the board

    With BRIEFKIN_TOKEN set, the agent sends the result as one card that updates in place every run. Without it, you get the report in the chat.

Install and run it

Pick your agent, where the data comes from and how to schedule it; the commands below follow.

Agent
Data
Schedule
  1. Install the skill

    One command. It puts the SKILL.md in ~/.claude/skills and in ~/.agents/skills, which Codex, Cursor, Gemini CLI, Copilot CLI and OpenCode read too.

    shellcurl -fsSL https://briefkin.dev/skills/posthog-experiment-results/install | sh
  2. Run it

    Ask Claude Code:

    promptRun the posthog-experiment-results skill

    Or paste the prompt instead of installing:

    promptUsing PostHog (the MCP server's experiment tools, or the experiments API), list my running experiments and get each one's results. For every variant give exposures, conversion on the primary metric, and the chance to win (or p-value if the experiment is frequentist), plus days running. Flag experiments with a clear winner, experiments over 28 days without a result, and exposure splits more than 10% off plan. End with one line per experiment: ship, end, or keep running. Do not change any experiment.
  3. Put it on a schedule

    Mondays at 9:00, from cron on any machine where Claude Code is signed in:

    crontab0 9 * * 1 claude -p "Run the posthog-experiment-results skill"

    Add it with crontab -e. A laptop asleep at that time runs it when it wakes if you use launchd or Task Scheduler instead; GitHub Actions works for the API route.

  4. Send it to your board

    Connect Briefkin once with curl -fsSL https://briefkin.dev/install | sh and every run lands on your board as a report card. How setup works

View the SKILL.mdRaw file
SKILL.md---
name: posthog-experiment-results
description: "Every Monday: each running PostHog experiment's variants, exposures, conversion and chance to win, how long it has run, and which are ready to call. Use when the user asks for this, or when it runs on a schedule."
---

# PostHog experiment results

Every Monday: each running PostHog experiment's variants, exposures, conversion and chance to win, how long it has run, and which are ready to call.

Experiments get launched with energy and then forgotten. The one that reached a clear result two weeks ago keeps holding half your users on the old version, and the one that will never reach significance keeps running because nobody looked.

Every Monday the agent reads each running experiment's results from PostHog and posts one table: variants, exposures, conversion on the primary metric, the chance to win (or p-value), and days running. Product managers and engineers see which experiments to ship, which to end, and which need more time.

## Two ways to get the data

Use the API or MCP route when you have an MCP server, a CLI or an API token for the app. Otherwise, if you can drive a browser the user is signed in to (Claude in Chrome, a Playwright or Chrome DevTools MCP server), use the browser route. Say which one you used.

## Route A: API or MCP server

### What you need

- The PostHog MCP server (`https://mcp.posthog.com/mcp`), or a personal API key with `experiment:read` in `POSTHOG_API_KEY` for the REST calls.
- The project ID and your PostHog host (US or EU cloud, or your own).

If something is missing, say what and stop; do not invent numbers.

### Steps

1. **List running experiments** (PostHog): Call `GET /api/projects/:project_id/experiments/` (or `experiment-list` on the PostHog MCP server) and keep the experiments with a `start_date` and no `end_date`, which is what PostHog shows as running. Note `name`, `feature_flag_key`, `start_date` and the primary `metrics`.
2. **Get the results** (PostHog): On the PostHog MCP server, call `experiment-results-get` for each experiment (`experiment-metrics-recalculation-latest-retrieve` returns the last computed run without starting a new one). Over REST, start a run with `POST /api/projects/:project_id/experiments/:id/metrics_recalculation/` and poll `GET .../metrics_recalculation/:recalculation_id/` until its `results` are complete.
3. **Read each experiment**: Per variant: exposures, conversion on the primary metric, and the statistic the experiment uses (chance to win for Bayesian, p-value and interval for frequentist). Flag a winner, an experiment over 28 days without a result, and exposures more than 10% away from the planned split.
4. **Post the readout**: One row per variant, experiments with a decision first, and one line per experiment on what to do: ship, end, or keep running.

## Route B: the browser

### What you need

- Signed in to PostHog with at least member access to the project; the experiments page needs the Experiments product enabled for the project.

If something is missing, say what and stop; do not invent numbers.

### Steps

Start at https://us.posthog.com/experiments. The user is signed in already; never type passwords or codes, and ask if a sign-in page appears.

1. **Open the experiments list**: Open `https://us.posthog.com/experiments` (or `https://eu.posthog.com/experiments` on EU cloud); PostHog opens your current project. Set the **Status** filter to **Running** and note each experiment's name and start date.
2. **Read each experiment**: Open each experiment. The results section shows one row per variant with exposures, the primary metric's value, and the credible or confidence interval; Bayesian experiments show **Win probability** and frequentist ones show a p-value. Days running is the time since the start date.
3. **Check the split**: Compare exposures across variants; flag a variant more than 10% away from the planned split, which PostHog also warns about above the results.
4. **Decide per experiment**: Mark a clear winner, flag anything running over 28 days without a significant result, and write one line per experiment: ship, end, or keep running. Do not click **Ship a variant** or **End experiment**.

Note: The experiment page is the better route here: it already shows the same results the API computes, with PostHog's own significance warnings. Results load on demand, so allow a few seconds per experiment, and a recalculation may run when you open one.

## The report

- **Running experiments** and how many have a result.
- **Per variant**: exposures, conversion, chance to win or p-value, days running.
- **Flags**: winners, experiments running long without a result, uneven exposure splits.
- One line per experiment: ship, end, or keep running.

## Post it to Briefkin

When `BRIEFKIN_TOKEN` is set, or `~/.config/briefkin/env` exists (the installer from https://briefkin.dev/docs writes it), also send the result to the user's Briefkin board as one card. Keep the `item_key` the same on every run so the card updates in place, use a new `event_id` per run, and put today's real numbers in place of the example ones. If the request fails, say so once and carry on.

```bash
[ -n "${BRIEFKIN_TOKEN:-}" ] || . ~/.config/briefkin/env
curl -sS -X POST "${BRIEFKIN_URL:-https://briefkin.dev}/v1/events" \
  -H "authorization: Bearer $BRIEFKIN_TOKEN" \
  -H "content-type: application/json" \
  -d @- <<'JSON'
{
  "v": 1,
  "event_id": "posthog-experiment-results/<date of this run, e.g. 2026-10-16>",
  "item_key": "posthog-experiment-results",
  "project": "Acme",
  "source": "PostHog experiment results skill",
  "kind": "update",
  "title": "1 of 4 experiments has a winner",
  "summary": "4 experiments are running. **New pricing page** has a **96%** chance to beat control on checkout started. **Onboarding checklist** has run 31 days without a clear result.",
  "widget": [
    {
      "type": "Table",
      "props": {
        "columns": [
          "Experiment",
          "Variant",
          "Exposures",
          "Conversion",
          "Chance to win",
          "Days"
        ],
        "align": [
          "left",
          "left",
          "right",
          "right",
          "right",
          "right"
        ],
        "rows": [
          [
            "New pricing page",
            "test",
            4210,
            "6.8%",
            "96%",
            18
          ],
          [
            "New pricing page",
            "control",
            4185,
            "5.9%",
            "4%",
            18
          ],
          [
            "Onboarding checklist",
            "test",
            2940,
            "41.2%",
            "58%",
            31
          ],
          [
            "Onboarding checklist",
            "control",
            2988,
            "40.6%",
            "42%",
            31
          ],
          [
            "Annual plan nudge",
            "test",
            1120,
            "3.1%",
            "71%",
            9
          ],
          [
            "Annual plan nudge",
            "control",
            1098,
            "2.6%",
            "29%",
            9
          ],
          [
            "Shorter signup form",
            "test",
            610,
            "22.0%",
            "64%",
            3
          ],
          [
            "Shorter signup form",
            "control",
            598,
            "21.1%",
            "36%",
            3
          ]
        ]
      }
    },
    {
      "type": "Callout",
      "props": {
        "text": "Ship the new pricing page. End the onboarding checklist test: after 31 days the gap is 0.6 points and still undecided.",
        "tone": "ok"
      }
    }
  ]
}
JSON
```

Widgets come from the Briefkin catalog (https://briefkin.dev/v1/catalog). Without a token, just give the user the report.

## Run it on a schedule

Suggested: Mondays at 9:00 (`0 9 * * 1`). From cron, with the agent signed in on that machine:

```bash
0 9 * * 1 claude -p "Run the posthog-experiment-results skill"
# Codex: codex exec, Gemini CLI: gemini -p, Copilot CLI: copilot -p, OpenCode: opencode run
```

Or use the agent's own scheduler (Claude Code routines, Codex automations, Claude or ChatGPT scheduled tasks): https://briefkin.dev/skills/posthog-experiment-results#schedule

Source: https://briefkin.dev/skills/posthog-experiment-results

What your agent needs

  • The PostHog MCP server (https://mcp.posthog.com/mcp), or a personal API key with experiment:read in POSTHOG_API_KEY for the REST calls.

  • The project ID and your PostHog host (US or EU cloud, or your own).

Connecting PostHog

  1. PostHog MCP server, PostHog's official remote server at https://mcp.posthog.com/mcp. Sign in with OAuth, or let the PostHog Wizard install it in your client; it routes to the US or EU cloud based on your account. Clients without OAuth can send a personal API key made with the MCP Server preset as a Bearer token (docs).
  2. Query API with a personal API key. Create one in your PostHog account settings, store it in POSTHOG_API_KEY with POSTHOG_PROJECT_ID and POSTHOG_HOST (https://us.posthog.com or https://eu.posthog.com), and give it the query:read scope, plus insight:read or feature_flag:read if the report uses them.

Add ?readonly=true to the MCP URL (or send x-posthog-read-only: true) to drop every tool that creates, updates or deletes; on the API, read scopes alone do the same.

Questions

Bayesian or frequentist?

PostHog lets each experiment use either method, and the skill reports whichever the experiment is set to: chance to win for Bayesian, a p-value and confidence interval for frequentist. The column is labeled accordingly, so two experiments on different methods are not read the same way.

Will it end or ship an experiment?

No. The PostHog MCP server has tools to end experiments and ship variants, but this skill only reads results and recommends. Shipping changes the feature flag for real users, so that stays a decision you make in PostHog.

Do I need Briefkin to use the PostHog experiment results skill?

No. The skill runs with your own agent and gives you the report in the chat. With a Briefkin token set, each run also posts it to your board as a report card, so you can glance at it instead of asking again.

Which agents can run it?

Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot CLI, OpenCode, Windsurf and Amp all read SKILL.md files. The install command puts it in ~/.claude/skills and ~/.agents/skills, where they look. ChatGPT, the Claude app and Grok take the prompt instead. Pick your agent under "Install and run it" and the commands change to match.

Can it work without an API or MCP server, from my browser?

Yes. Switch the data source to "Your browser" and the steps change to the PostHog pages themselves: the agent opens us.posthog.com/experiments in a browser you are signed in to and reads the numbers off the page. Claude Code, Codex, Cursor, Claude app and ChatGPT can drive a browser out of the box; Gemini CLI, Copilot CLI and OpenCode need a Playwright or Chrome DevTools MCP server first. It is slower than the API and needs you signed in, so it suits a report you run by hand or from a scheduler on your own computer.

Read this report on your board, not in a chat log

Skills are free to use with any agent. Briefkin is where their results land: one board for every report, number and question your agents send, with a ping when something needs you.

Free during the open beta. No card.