This repository is public technical documentation for SplitLaunch. It is meant to help AI agents discover and operate the SplitLaunch API safely.
- Read SKILLS.md.
- Prefer the live command contract at
https://www.splitlaunch.dev/api/skillswhen you need current details. - Use QUICKSTART.md for install and first experiment setup.
- Use skills/README.md for capability-specific pages.
- Use runbooks/README.md for repeatable workflows.
SplitLaunch is package-first A/B testing for customer-owned pages. The user keeps their own website. SplitLaunch provides a package, CLI, browser pixel, command API, conversion tracking, and reporting API.
- Keep
sl_live_...API keys out of browser code and GitHub commits. - Browser code should receive only public
px_...pixel IDs. - Use
https://www.splitlaunch.devfor API calls. - Resolve the correct
projectIdbefore write commands. - Allowlist domains before redirecting traffic.
- Create experiments in draft, then ask before setting them to
running. - Track conversions only after the successful user action.
- Read back resources after writes.
- Explain results with sample size, conversion rate, lift, revenue, channel context, and uncertainty.
POST https://www.splitlaunch.dev/api/external/v1/commands
Authorization: Bearer sl_live_REPLACE_WITH_REAL_KEY
Content-Type: application/json
Request envelope:
{
"command": "projects.get",
"input": {}
}projects.getprojects.listdomains.addexperiments.createexperiments.set_statusgoals.createreports.experiment.detailsreports.experiment.channelsusage.summary
This repo does not contain the private SplitLaunch app source code. Do not look for internal database migrations, billing code, dashboard implementation, or pixel runtime source here. Use the hosted API and docs links.