1. Identify and track
Two typed calls from your app, across every product you run. Backfill a year of history without emailing a year of users.
Cowliss
Identify your users, track what they do, segment them, then write the emails that follow in TypeScript. Cowliss keeps the profiles and the waits that last for weeks.
2. Group them
Segments are rules over traits and events, recomputed on every write. Entering one can start a journey.
3. Write the journey
defineJourney({ trigger, from, run }) in your repo. One execution per user, waits that survive a restart, events delivered as they arrive.
Hand it to an agent
Anything the dashboard does, the cow CLI and the MCP server do too.
Read these docs as an agent
Section titled “Read these docs as an agent”Every page is markdown too: append index.md to its URL (/journeys/index.md). The index is /llms.txt, the whole site is /llms-full.txt, and the API contract is /openapi.json.
QuickstartFrom nothing to a journey that sends a real email.
ProjectsJourneys and emails as code in your own repo.
API referenceGenerated from the same zod schemas the API validates with.
CLI referenceGenerated from the Cowliss CLI's own command definitions.
CLI and MCP for agentsEverything the dashboard does, without the dashboard.