Stop guessing.
Start testing.
Scale what works.
The agent-first experimentation platform for product and marketing teams. Your agent drafts the test, a person approves it, and Causalo proves what works.
- Data hosted in the EU
- A person approves every start
- Agents can draft, never launch
Your agent does the work. You make the call.
Causalo works from the CLI, in the coding agent you already use. Your agent drafts the experiment and links its pull request; you approve it and read the result. MCP is coming.
›
Example result
+8.4%
paid conversions vs. control
REAL CHANGES, MEASURED WHERE THEY HAPPEN
One loop, built to get smarter with every experiment.
Deciding what to test, checking it is allowed, testing it properly and remembering what you learned usually takes four teams and a quarter. Causalo is being built to make it one loop. The testing runs today; the rest is in development.
01 / 04
Hypothesize
Hypothesize
Your team and your agents propose what to test. Ranking ideas by expected value and risk is in development.
H-052
Paywall after the first saved search
The research behind the learning loop.
The learning loop is what we are building next, and it starts from published research, not from results of our own. Language models can rank which interventions will have larger effects: across 70 survey experiments their predictions tracked the real ones about as well as expert forecasters, while overstating the size (Ashokkumar, Hewitt et al., Nature, 2026). On real marketing A/B tests, the same kind of prediction is weak out of the box and becomes useful only once it is calibrated on a company's own past experiments (Hut & Masoero, Amazon, COLM 2026). That sets the order. Today Causalo measures experiments properly, starting on our own products. The loop, once built, will use those measured results to rank the next ideas.
What science already knows
The loop will start from randomised results in behavioural science, economics, marketing and published A/B tests.
What your stack shows
Google Search Console today. Analytics, ads, CRM and subscriptions next.
What your own experiments proved
Every measured result is kept with its evidence, ready for the loop to learn from.
Coming · suggested test idea
Idea: show the saving in kroner on the sign-up button.
Ranked by expected lift, once the loop is built.