Skip to main content
Agent-first product analytics

Let your agent measure what it ships.

Product analytics for teams moving the build-measure-decide loop into coding agents. No SQL detours.

Setup guides for
  • Claude Code
  • Claude Desktop
  • Codex
  • Cursor
  • Custom MCP
One measurement loop

Ask. Measure. Decide.

Poolstatis gives your coding agent the metrics and decision rules your team approved.

Open quickstart
You askDid onboarding improve?
Poolstatis MCPquery_funnel activation
Agent returnsEvidence, limits, next action.
For product teams

See growth, not event lists.

See the lift, the drop-off, and the next move at a glance.

LiftFunnelsCohortsNext move

Start with one real question.

Create a project, send evidence, and ask from the coding workspace you already use.

Open MCP reference
01

Create

Sign up, verify your email, and create the first project.

A project and scoped keys are ready.
02

Prove

Send a real event and activate a purpose-first metric.

Accepted evidence is queryable.
03

Ask

Connect MCP and ask the product question in your coding workspace.

The answer stays tied to registered meaning.
ResultYour agent can query only the metrics and funnels your team registered.

A small analytics core.

The SDK records facts. The registry supplies meaning. MCP returns bounded answers.

EventImmutable fact
EntityMutable state
MetricPurpose attached
FunnelGoal attached
registry

Purpose-first metrics

Every metric states the decision it supports before an agent can use it.

query

Typed analytics surface

Trends, funnels, retention, lifecycle, and stickiness stay inside a narrow DSL.

data

Drift stays visible

Events are registered or flagged wild, so agents can audit the source.

infra

Self-hostable core

Postgres, clean store interface, MCP server, and SDK stay in your environment.

Source available. Cloud beta hosted.

Run the agent-native analytics loop in your environment, or use hosted signup if you want Poolstatis to operate the platform.

Access model

Choose where the analytics loop runs.

Cloud is the same agent-native analytics contract, hosted for teams that do not want platform ops in their backlog.

Source coreAvailablePolyForm Shield
CloudBeta signupVerified accounts
AreaSelf-hosted corePoolstatis Cloud
RuntimeAPI, admin, MCP, SDKManaged workspace
DataYour PostgresHosted ingest and retention
Keyspk_ and pt_ generated by youOne-time setup tokens
OpsYour uptime and backupsAuth, backups, upgrades

Cloud, without the platform work.

Sign up, verify your email, and create a project. Scoped setup tokens appear once.

MCP-ready workspaceHosted tokens and client presets without running the platform yourself.
Managed event storeManaged ingest, registry, and query APIs for agent verification.
Human admin authHuman sign-in stays separate from scoped runtime keys.
Cloud waitlist

Want Cloud updates? Join the waitlist. Beta signup is already open.

Questions before you start.

Is this just PostHog? +
PostHog also offers broad product analytics and MCP access. Poolstatis is intentionally narrower: every metric needs a purpose and every funnel needs a goal. Coding agents get typed queries, while people keep a reviewable workspace. It does not claim broader feature parity.
Do I have to write SQL? +
Not for normal product analysis. Your agent uses typed queries over registered metrics for trends, funnels, retention, lifecycle, and stickiness. Advanced operators can still inspect the underlying system when needed.
Can product teams use it too? +
Yes. Agents handle instrumentation and typed queries, while product people review funnels, trends, retention, lifecycle, people, and data quality in a human workspace.
Where does my data live? +
Self-hosted Poolstatis keeps data in your Postgres. Poolstatis Cloud runs the same contract as a managed beta service, with hosted human auth separate from scoped runtime keys.
Which agent works with it now? +
Poolstatis ships MCP templates for Claude Code, Claude Desktop, Codex, Cursor, and custom MCP hosts. The server uses standard stdio MCP values: command, args, and env.

Measure the release you ship.

Start in Cloud or run the source-available core in your own environment.