PostHog + AgentCeres

What AgentCeres reads from PostHog

• Your PostHog projects, so AgentCeres reads from the right one
• The dashboards your team has built in each project — their names and descriptions
• Experiments — the A/B tests set up in a project, and when they ran
• Surveys — the in-product surveys configured in a project, with their questions
• Session recordings — the list of recorded user sessions in a project
• Actions — the named events and page views your team has defined
• Feature-flag status — whether a given flag is active or stale

How AgentCeres uses PostHog

Once you connect PostHog, your agents can read what your project holds: the dashboards your team has built, your experiments and surveys, the list of recorded sessions, the actions you've defined, and whether a feature flag is active. This connection is strictly read-only: AgentCeres only reads from PostHog. It never creates events, edits dashboards, flips feature flags, or changes anything in your project. It looks, it learns, and it leaves your analytics exactly as it found them.

Your AI Growth Officer is the one that reads it — for your daily growth brief and whenever a question calls for it: to see which experiments and surveys are running, which actions your team treats as key moments, and whether a feature flag is live, so its recommendations start from how your product is actually set up. No specialist on your team is built around product analytics. Every finding is tied back to the specific PostHog project and item it came from.

What AgentCeres reads here is your project's setup and lists, not your event data: it does not run funnel, retention, or trend queries against your events, so it does not report where users drop off. The value is that your agents can check what you already track and test themselves, instead of asking you to open PostHog and describe it. You get the read; you keep full control of your project.

In practice