Can it change feature flags?

It reads them, which is what you want here. A flag is a production switch, and the value of having an agent look at flags is correlating a rollout with a change in behaviour, not flipping one because the numbers looked good.

Will it invent an explanation if there is not one?

It can, which is the honest risk with any agent asked "why". The mitigation is in how you ask: request the evidence alongside the claim, and treat an answer with no query behind it as an opinion. A cause that names the events it read is checkable; one that does not is a guess.

Does it work with self-hosted PostHog?

The connector talks to the PostHog API, so a self-hosted instance works provided the agent can reach it and you authorise it the same way. What it reads is whatever your own project permissions reach, no more.

Analytics

PostHog + Burrak AI

Ask what the numbers mean, not just what they are.

Connects with OAuth — revoke any time

PostHog will tell you a funnel dropped without telling you why, and the gap between those two things is usually an afternoon of clicking. An agent can query the insights, look at the events underneath, check whether a flag rollout lines up with the change, and come back with a short answer that names a cause you can go and check — rather than another chart.

In the connector catalogue: Query insights, inspect events, check feature flags.

Things you could ask for

Written the way you would actually say them. An agent plans the steps itself.

“Signups fell this week. Look at the funnel and tell me at which step, and for which platform.”
“Did the feature flag we rolled out on Tuesday change activation at all?”
“Which events fire most often for people who stay past week two but not for people who leave?”
“Summarise how this release changed behaviour, in terms a non-analyst would understand.”
“Every Monday, tell me what moved last week and what plausibly caused it.”

An agent reading analytics will offer causes, and a cause it offers is a hypothesis, not a finding. Ask it to say what would confirm or kill each one — an explanation you cannot test is worse than the chart you started with.

Connecting it

1Create a Burrak account — the free plan is enough to try this.
2Open Connectors and choose PostHog.
3Authorise Burrak in PostHog's own window. Your password never reaches us; we receive a token you can revoke whenever you like.
4Give an agent a task. It will use the connector when the work calls for it.

Start free How connectors work

PostHog and Burrak, in practice

Can it change feature flags?: It reads them, which is what you want here. A flag is a production switch, and the value of having an agent look at flags is correlating a rollout with a change in behaviour, not flipping one because the numbers looked good.
Will it invent an explanation if there is not one?: It can, which is the honest risk with any agent asked "why". The mitigation is in how you ask: request the evidence alongside the claim, and treat an answer with no query behind it as an opinion. A cause that names the events it read is checkable; one that does not is a guess.
Does it work with self-hosted PostHog?: The connector talks to the PostHog API, so a self-hosted instance works provided the agent can reach it and you authorise it the same way. What it reads is whatever your own project permissions reach, no more.