AI agent · Cee
Cee is Countly’s AI agent. It works on your live behavioral and profile data: it answers questions, builds cohorts and funnels, explains changes and drafts the next action. When you self-host, Cee’s product agents can run fully inside your infrastructure.
What you can ask
Every answer shows the query and the data Cee used. Cohorts, alerts and messages are drafted for review, never sent on their own.
“Why did signups drop last week?”
Compares the period, finds the segment that fell (Android 5.2, day of the release), shows the chart and the query.
“Create a cohort of users who ran a drill query 10+ times in two weeks.”
Builds and saves the cohort, ready for a message or an experiment.
“Build a funnel from signup to first purchase and show where users drop.”
Builds the funnel, highlights the biggest drop and shows the query.
“Which features do churned users stop using first?”
Runs the analysis across events and profiles and writes a finding with the sequence.
Capabilities
Cee reads the same schema your team governs in the Data Manager, so it speaks in your event names and segments, not in generic metrics.
Ask why signups dropped last week and get the segments, the release and the chart that explain it, with the query it ran.
Cee creates cohorts, funnels, drill queries and dashboards from a sentence. Review, adjust, save.
Cee watches key metrics, flags anomalies and writes a finding with probable causes before anyone opens a report.
From a finding to a message, an experiment or a webhook, Cee drafts it inside the same system.
Want your own AI to use this context? Connect your model, copilot or agent to Countly through MCP.
Cee’s product agents for Cohorts, Funnels, Journeys, Drill and Drill Insights can run fully inside your infrastructure when configured with a self-hosted inference endpoint or your own AI provider. The Documentation Agent, the Countly AI Gateway (EU) and third-party AI providers run outside.
AI Analytics
AI Analytics tracks the adoption, cost, quality and performance of the AI features in your own product. Analyze models, agent tools and conversations next to the rest of your product data, so the AI you build is measured like everything else.
In depth
The parts that matter when you evaluate this area seriously: what it does, how it behaves at the edges, and what your team controls.
Explore
Cee turns a question into a drill query, shows its work, and lets you pin the result to a dashboard or save it as a finding. Analysts move faster, product managers stop waiting for analysts.
Governance
Cee inherits the permissions of the person asking. It reads only the apps and data that person can read, and it drafts rather than executes, so cohorts, alerts, messages and experiments are created in a pending state and enabled by a human. Every request, query and draft is logged for review.
How it works
Cee is a feature of the platform, not a separate SaaS. It inherits your permissions and your schema, and its product agents can run where your data runs.
Type a question in Drill, on a dashboard or in the Cee panel. Cee understands your event and segment names.
Cee shows the query, the data range and the segments it used. Adjust and re-run in one click.
Pin the chart, save the cohort, create the funnel. Everything lands in the same workspace.
Turn the finding into a message, an experiment or a webhook, or save it for the team.
FAQ
Straight answers on how this part of Countly works in practice. For anything else, talk to us.
By default Cee runs through the Countly AI Gateway (EU) or a third-party provider; on Enterprise, the product agents can use your own endpoint.
Cee’s product agents can run fully inside your infrastructure when configured. The Documentation Agent, the Countly AI Gateway (EU) and third-party AI providers run outside.
Cee works on the governed data layer, masked, hashed and consent-filtered. It never sees raw PII that your masking rules removed at ingestion.
No. Cee drafts cohorts, messages and experiments; a person with the right permission reviews and enables them. Every action is logged.
Yes. The API and MCP expose real-time behavioral and profile context, so your agents and product logic work on the same live data.
