Agent setup
Pixel installed in customer repo
SplitLaunch CLI • pixel connected
A/B testing your agent can install and run.
Install one package, connect your API key, and let your AI agent create experiments, track conversions, and retrieve results through the API.
runningexp_42
Control URL50%
Variation URL50%
Sticky assignmenton
splitlaunch init
$ npm install @splitlaunch/ab
$ npx splitlaunch init
Paste your SplitLaunch API key: sl_live_...
Validated account.
Pixel ID: px_7H4K92
Agent docs: https://splitlaunch.dev/api/skillsagent command
{
"command": "experiments.create",
"trafficAllocation": 50,
"goal": "Lead"
}A lightweight testing layer for agent-led teams.
SplitLaunch is designed for users who want experiments running in their own codebase without enterprise platform complexity.
installed in the user's repository
1
package
events available on the first plan direction
100k
+
traffic allocation configured by API
50
/50
pixel collection after install
24
/7
Built for teams shipping on modern web stacks
AI agencies
SaaS teams
Shopify builds
Next.js apps
Growth teams
Product agents
Landing pages
Custom sites
The experiment layer your agent can control.
SplitLaunch keeps the product simple on purpose: real pixel behavior, a focused API, and agent-readable skills instead of a bloated experiment dashboard.
Agent-created experiments
Define control and variation URLs, traffic allocation, targeting filters, and launch status through an agent-friendly command API.
The dashboard stays small because the agent is the main interface.
95
85
90
Pixel assignment and redirects
Assign visitors deterministically, preserve sticky cookies, and redirect traffic to the assigned URL with parameters intact.
Experiment live
Control and variation URLs launched
Agent API • traffic split active
Conversion tracked
Lead event attributed to variation
Browser pixel • event collected
Rich event collection
Capture sessions, exposures, referrers, UTMs, click IDs, device context, viewport, geo context, events, and conversion metadata.
Custom conversion events
Track leads, trials, purchases, demos, or any custom action directly from the user's app.
Agents can document and wire conversion events after setup.
Your site stays your site. SplitLaunch becomes the experiment layer around it.
The package installs the pixel, the API controls the test, and the agent retrieves performance when the user asks for results.
Everything needed. Nothing that turns it into a heavy dashboard.
CLI setup
The terminal asks for the API key and retrieves the pixel ID.
Public pixel
Browser code only receives a safe public pixel identifier.
URL experiments
Control URL, variation URL, split, status, and targeting.
Agent API
Commands for tests, goals, usage, billing, and reporting.
Usage billing
Plans can be tied to events, projects, and usage limits.
Simple dashboard
Credentials, install command, billing, usage, and docs only.
“The product idea is simple: install the package, let the agent create the experiment, and get the result through the API.”
Plans for teams that just want tests running.
Final limits can change later. The pricing direction is recurring, simple, and tied to usage instead of enterprise complexity.
Launch
For builders running agent-led experiments on one project.
Scale
For agencies and teams shipping experiments repeatedly.