Skip to content

llms.txt

llms.txt is a plain-text, machine-readable guide that tells an AI coding agent exactly how to add Versuch to a project. Point your agent at it and it can install the tracker, wire up conversions, and set up experiments and flags without you writing the boilerplate.

https://versuch.ai/llms.txt

Give the URL to your agent (Claude Code, Cursor, Copilot, …) or paste the contents into a task. Fetch it fresh each time so you get the current, verified snippets.

The file is a step-by-step integration guide covering:

  • Install the tracking snippet: the exact <script> tag (loader https://cdn.versuch.ai/v.js, data-site-id, data-api) plus framework-specific placement for plain HTML, Next.js (App and Pages Router), Vite/React/Vue, Astro, SvelteKit, Nuxt, and Remix.
  • Track conversions: declarative data-versuch-event / data-versuch-props attributes and the programmatic versuch.track(name, props?) API.
  • Identify users: versuch.identify(userId) after sign-in and versuch.reset() on sign-out.
  • Verify: how to confirm events are arriving.
  • A/B experiments: how to create and manage them via the MCP server or the REST API with the project secret key.
  • Stripe revenue attribution: forwarding the versuch_visitor_id / versuch_session_id cookies as Checkout metadata plus the email fallback, so purchases land in the buyer’s journey and per-experiment results.
  • Feature flags: reading flag values with @versuch/sdk (browser and server) and creating/toggling flags via the API or MCP.
  • Reference: the loader URL, ingest API, the window.versuch public API, and the declarative attributes.

Paste this into your coding agent:

Read https://versuch.ai/llms.txt and follow it to add Versuch to this project. My public ingest key is vpk_live_….

The agent adds the script tag to your layout, wires conversion events onto your CTAs, and (if you provide the secret key or connect the MCP server) can go on to create experiments and flags.