AI agents for PostHog
A Monmo agent works in PostHog the way an engineer who cares about product data would. It watches session replays where users struggle, writes SQL insights for questions dashboards do not answer, reads experiment and survey results, drafts bugs from error tracking, and prepares feature flag rollouts. It uses your own PostHog API key and asks before it changes a flag, an experiment or a dashboard.
Updated 2026-11-13 · Free to start. Paid plans from $15 a month.
What agents do in PostHog
Watch replays where users get stuck
It filters session replays to people who dropped at a funnel step or hit an error, watches a sample, and writes what they tried to do, with timestamps for the moments worth seeing yourself.
AskWatch 20 replays of users who abandoned checkout this week and tell me what went wrong.
Write a SQL insight
For questions the standard insights cannot answer, it writes a SQL query in PostHog, runs it, explains the joins in plain words and adds it to a dashboard once you approve.
AskWrite a SQL insight for weekly active teams with at least three members, and explain how it counts.
Prepare a feature flag rollout
It drafts the rollout plan for a flag, such as 10 percent of users, then 50, then everyone, with the metrics to watch at each step. Every change to the flag asks first.
AskPlan the rollout of the new-editor flag in three steps and tell me what to watch before each one.
Read an experiment
When an experiment reaches its planned length, it reads the results, checks the goal metric and the secondary metrics, and writes a short note with the decision left to you.
AskIs the pricing page experiment done? Summarise the result and what you'd do.
Summarise survey answers
It reads the open answers from a PostHog survey, groups them into themes with a count for each and quotes the clearest examples, so the team reads one page instead of every reply.
AskGroup the answers to the cancellation survey into themes and quote two examples of each.
Turn errors into issues
From error tracking it groups new exceptions, links each to the replays where it happened, and drafts a GitHub or Linear issue with the steps, stack trace and affected users.
AskDraft Linear issues for this week's top five new errors, with a replay link for each.
Daily product health note
Each morning it checks signups, active users and errors against yesterday, and drafts a three-line note for Slack that points out anything unusual.
AskEvery weekday at 9, draft a three-line product health note for the engineering channel from PostHog.
How it connects
- 1You connect PostHog with a personal API key you create in your PostHog settings, choosing its scopes and projects. Say whether your project is on the US or EU cloud, or self-hosted, so the agent uses the right address.
- 2You choose which agents may use PostHog. A web developer or data analyst agent fits; a sales agent needs only the summaries.
- 3Watching replays, running insights and drafting issues need no approval. Changing feature flags, starting or stopping experiments, launching surveys, saving dashboards and creating issues ask first.
More in Analytics and data
Other apps of this kind that agents work in, each with its own page.
Questions people ask
Can an AI agent watch PostHog session replays?
Yes. A Monmo AI agent filters replays to the users and moments you care about, watches a sample and writes what happened, with timestamps for the parts worth seeing. It works through the PostHog connection, or in its own browser when the connection does not cover something. Replays can show personal data, so give access only to the agents that need it.
Can it change feature flags in PostHog?
It prepares flag changes and makes them only after you approve. The card shows the flag, the old and new rollout, and the metrics it will watch. Turning a flag on for everyone is a production change, so it always waits for you, and afterwards the agent can watch errors and tell you if anything rises.
What scopes should the PostHog API key have?
The least the job needs. For reports and replays, read scopes on the projects you choose are enough. Add write scopes for feature flags, experiments or dashboards only if you want the agent to make those changes after your approval. You can revoke the key in PostHog at any time.
Can it file bugs from PostHog errors?
Yes. The agent groups new exceptions from PostHog error tracking, links each to the replays where it happened and drafts an issue in GitHub or Linear with steps, stack trace and the number of users affected. Creating the issues asks first, and one card can cover the week's list.