Realtime personalization and product analytics, set up by your coding agent.

Know what each user is doing right now and act on it within the same request. Runs on Cloudflare's edge, with one profile per user.

A live profile and the decision it just madeExample data

user u_4821

One profile across the website and the app

Plan
Trial, day 9
Account
Acme Studio
Sessions, 7 days
6
Last seen
14:02:11

Segments

  • Trial users, high intent
  • Design plan prospects

Events, last 5 seconds

  1. 14:02:07webpricing_page_viewed
  2. 14:02:09appdocument_created
  3. 14:02:10appshare_dialog_opened
  4. 14:02:11profilesegment_enteredTrial users, high intent

Decision for slot upgrade_banner

Variant high_intent

This user is in high_intent_trial: 1 pricing view in the last 7 days, the threshold. The rule matched.

Trial users, high intent include users where all are true

  1. account.planistrial
  2. pricing_page_viewedat least1 timeinlast 7 days
Rule
r1When in segment high_intent_trial, serve high_intent.
Freshness
FreshThe profile has applied the 14:02:07 pricing view.
Path
ProfileRead from this user's Durable Object.
Decided
14:02:11In the first HTML, before the page is sent.
Props
{ text: "Your trial ends in 5 days. Compare plans." }

No model call and no warehouse query in this path.

Two products, one profile

Marketing and product teams work from the same user

UserCurrent Web and UserCurrent App run on one shared core. Turn on both for a project and what people do on the site and in the product land on one profile, with no data copy and no identity stitching.

For growth and marketing

UserCurrent Web

Personalize the site, run web experiments and trigger journeys in realtime, with no CMS migration.

  • Experiences and slots on your own pages
  • Acquisition analytics and web experiments
  • Outbound messages and journeys
About UserCurrent Web

For product and engineering

UserCurrent App

Flags, experiments and product analytics on one profile, set up from the repo by your agent.

  • Feature flags with percentage and behavioral targeting
  • Product experiments and feature usage
  • Funnels, paths and behavior over time
About UserCurrent App
One profile, events from both productsExample data

user u_4821 Trial, Acme Studio

  1. Oct 24, 09:12webpricing_page_viewedutm_source: newsletter
  2. Oct 24, 09:14webcta_clickedlabel: Start trial
  3. Oct 24, 09:16appsigned_upmethod: google
  4. Oct 24, 09:16profileidentifyanonymous visitor joined user u_4821
  5. Oct 26, 15:40appdocument_createdkind: note
  6. Oct 28, 11:05webpricing_page_viewedutm_source: none
  7. Oct 28, 11:09appshare_dialog_opened
The same events, in analyticsExample data

Trial accounts that shared a document

31.4%

Last 7 days. +9.4 pts vs prior 7 days (22.0%).

0%10%20%30%40%Oct 15Oct 21Oct 28Share prompt launched34%

Why UserCurrent

Built so a decision can happen in the request

The design choices that make an in-request decision possible.

  • Realtime

    A behavior can change the next response. Profiles live in Durable Objects close to the user, so a decision needs no warehouse round trip.

  • Headless

    There is no CMS to adopt. Slots plug into your site and your CMS: UserCurrent decides what each slot shows, and your code renders it.

  • AI in the loop, out of the render path

    AI hooks write to the profile after events, so a page never waits on a model. Anything AI generates carries the AI label.

    AI

  • Guardrailed autonomy

    People set the budgets and limits and approve. The platform may run experiments and learn on its own inside them.

  • Analytics built in

    Product and acquisition analytics run on the same events that drive decisions, with no second pipeline to maintain.

  • One profile, every channel

    The web and server SDKs share one identity graph and one profile, so a visitor who signs in keeps their history. Mobile and desktop SDKs are next.

Agent-first setup

Your coding agent does the setup. The config lives in your repo.

Tracking plans, flags, segments, experiments and dashboards are a typed config file in your repository. Your agent edits it, the CLI validates it, and a person reviews the pull request.

Terminal
# Write a starter config for the repo
uc init

# Check the config, compile it and compare tracking calls with the plan
uc validate
uc scan

# See what would change, then publish it (a dry run until --apply)
uc diff
uc publish --environment dev
uc publish --environment dev --apply
Pull request from your coding agentExample data

Add a tracking plan and an upgrade banner

Agent
  • usercurrent.config.ts+38
  • src/lib/usercurrent.ts+12
  • src/layouts/Base.astro+4 −1
  • uc validate passed
  • uc diff: + segment high_intent_trial, + slot upgrade_banner
  • Agent keys cannot publish to Production. A person approves it.

The technology

Built on Cloudflare's edge. Your data stays in open tables.

Workers take events and decide, a Durable Object holds each live profile, and the event lake is open Iceberg tables on R2, not a proprietary store. Analytics read the lake directly.

Where it runs
  1. Your sites and apps

    • Web SDK
    • Server SDK
    • Astro integration

    Run wherever you host them

  2. Cloudflare

    • Edge Workers decide
    • Durable Objects hold live profiles
    • Lake Durable Objects write the event tables
    • R2 stores the event lake

    Close to every user, on one network

  3. Open Iceberg tables

    • Events
    • Sessions
    • Per-user daily rollups

    Read them with any Iceberg reader

Your site and apps call UserCurrent through its SDKs wherever they are hosted. See how it is built.

Data trust

Your data is never used for advertising or to train shared models.

  • Not for advertising

    Customer data runs your product and your experiences. It is never used for advertising.

  • Not for shared models

    Your data is never used to train shared models. AI hooks write to your profiles after events, and AI output is labeled as AI.

  • Yours to remove or take

    Deleting a person is built in. Per-person export and EU or US residency are partly built. Personal data is masked in agent answers.

Set up tracking with your coding agent

Start from a config file in your repo. The CLI and MCP server are part of the product, not an add-on.

Why this button?

It is this page's own suggestion for now. As you read, this site keeps a small profile of who each page is for and what it is about, in your browser, and picks this button from it. Nothing is sent.

See your vector on the live demo