Investigating Replay

Investigating a session recording

When a user asks "what happened in this session?" or provides a recording/session ID to investigate, gather all relevant context in parallel rather than making them ask for each piece.

Available tools

Workflow

Step 1 — Get recording metadata and person profile

Start with the recording to get metadata and the person's distinct ID:

posthog:session-recording-get
{
  "id": "<session_id>"
}

The recording id and the event $session_id are the same value. It selects the recording here and the same-session events in Step 2. The response includes distinct_id, person, start_time, end_time, duration, interaction counts, console error counts, and viewing status. Use the distinct_id to fetch the full person profile:

posthog:persons-retrieve
{
  "id": "<person_uuid_from_recording>"
}

Step 2 — Query same-session events

Use the recording id from Step 1 as the $session_id value. Get the timeline of what the user did during the session:

posthog:execute-sql
SELECT
    timestamp,
    event,
    properties.$current_url AS url,
    properties.$browser AS browser,
    properties.$os AS os,
    properties.$device_type AS device_type,
    properties.$screen_width AS screen_width
FROM events
WHERE $session_id = '<session_id>'
ORDER BY timestamp ASC
LIMIT 200

For sessions with many events, focus on the most informative ones:

posthog:execute-sql
SELECT
    timestamp,
    event,
    properties.$current_url AS url,
    if(event = '$exception', properties.$exception_values[1], null) AS exception_message,
    if(event = '$exception', properties.$exception_types[1], null) AS exception_type
FROM events
WHERE $session_id = '<session_id>'
    AND event IN ('$pageview', '$pageleave', '$autocapture', '$exception', '$rageclick')
ORDER BY timestamp ASC
LIMIT 100

No rows? Recover the event session ID

The recording id is the session ID. No rows means the session's events were ingested without it. Find candidates from the person's events in the recording window, padded by 100 seconds like the replay events query. person_id covers all of the person's distinct IDs:

posthog:execute-sql
SELECT
    properties.$session_id AS session_id,
    count() AS event_count,
    min(timestamp) AS first_seen,
    max(timestamp) AS last_seen
FROM events
WHERE person_id = '<person_uuid>'
    AND timestamp >= toDateTime('<start_time>') - INTERVAL 100 SECOND
    AND timestamp <= toDateTime('<end_time>') + INTERVAL 100 SECOND
    AND properties.$session_id IS NOT NULL
GROUP BY session_id
ORDER BY event_count DESC
LIMIT 10

Continue only when one session ID clearly matches. Use it for the Step 2 and Step 3 queries only. The replay URL and all Replay Vision calls take the recording id.

Step 3 — Check for linked error tracking issues

If the recording has console errors or exceptions, find related error tracking issues:

posthog:execute-sql
SELECT DISTINCT
    properties.$exception_fingerprint AS fingerprint,
    properties.$exception_types[1] AS type,
    properties.$exception_values[1] AS message,
    count() AS occurrences
FROM events
WHERE $session_id = '<session_id>'
    AND event = '$exception'
GROUP BY fingerprint, type, message
ORDER BY occurrences DESC
LIMIT 10

If fingerprints are found, search for the corresponding error tracking issues to provide links and status:

posthog:query-error-tracking-issues-list
{
  "searchQuery": "<exception_type or message>"
}

Step 4 — Synthesize the investigation

Present the findings as a coherent narrative:

1. Who — person properties (name, email, country, plan, etc.)
2. What — sequence of pages visited and key actions taken
3. Problems — exceptions, console errors, rage clicks, and their frequency
4. Related issues — linked error tracking issues with their status (active/resolved)
5. Context — session duration, device/browser, activity score

Optional: AI summary via Replay Vision

If the user wants a deeper analysis without reading through events manually, offer a Replay Vision summary. Follow "check-then-scan" — don't scan blindly, a scanner can only observe a given session once.

1. Check for an existing summary. A scheduled scanner may already have one:

   posthog:vision-observations-list
   {
     "session_id": "<session_id>"
   }
   

   Look for an observation where scanner_snapshot.scanner_type is summarizer and status is succeeded. If found, read scanner_result.model_output (title, summary, intent, outcome, friction_points, keywords) — done, no new scan needed.

2. Find a summarizer scanner if none exists yet:

   posthog:vision-scanners-list
   {
     "scanner_type": "summarizer"
   }
   

   • Exactly one → use it.
   • More than one → show the user the scanners (name + prompt) and ask which to use.
   • None → no summarizer scanner exists. See No summarizer scanner? Run a temporary one below.

3. Scan the session with the chosen scanner. Warn this is async and takes several minutes (rasterize + LLM):

   posthog:vision-scanners-scan-session
   {
     "id": "<scanner_id>",
     "session_id": "<session_id>"
   }
   

4. Retrieve the result by polling vision-observations-list (step 1) until the new observation reaches succeeded.

No summarizer scanner? Run a temporary one

If the project has no summarizer scanner, you can still produce a one-off summary with a throwaway scanner — but ask the user's permission before creating anything.

1. Ask permission to create a temporary summarizer scanner just to summarize this one session.

2. Create it disabled so it never sweeps on a schedule — a disabled scanner only runs when you trigger it on demand, so it won't touch other sessions or burn quota in the background:

   posthog:vision-scanners-create
   {
     "name": "Temporary on-demand summary",
     "scanner_type": "summarizer",
     "scanner_config": {
       "prompt": "Summarize what the user was trying to do, whether they succeeded, and any friction they hit."
     },
     "query": { "kind": "RecordingsQuery" },
     "model": "gemini-3-flash-preview",
     "enabled": false
   }
   

3. Scan this session on demand with the new scanner, then poll for the result:

   posthog:vision-scanners-scan-session
   {
     "id": "<new_scanner_id>",
     "session_id": "<session_id>"
   }
   

   Poll vision-observations-list until the observation reaches succeeded and read scanner_result.model_output.

4. Ask whether to keep or delete the scanner. Once you have the observation, ask the user if they want to keep the temporary scanner or delete it with vision-scanners-delete. Deleting is safe: the summary you just read is also emitted as an event that persists after the scanner is gone, so cleaning up the temporary scanner does not lose the result.

Tips

• Run steps 1-3 in parallel when possible — they're independent queries.
• If the recording has very few events, the session was likely very short. Note this rather than suggesting something is broken.
• Console error count from the recording metadata is a good signal for whether to dig into exceptions. If it's 0, skip step 3.
• The start_url from the recording tells you where the user's journey began — use this to frame the narrative.
• If person is null on the recording, the user was anonymous. Person properties won't be available, but events still are.

Related skills

• finding-sessions-to-watch — choose which sessions are worth investigating in the first place
• finding-replay-for-issue — start from an error tracking issue and find its linked recordings
• diagnosing-missing-recordings — when a recording that should exist doesn't
• creating-replay-vision-scanners — automate this kind of watching as a scheduled scanner