Quickstart

Get Tamda running in under 10 minutes. Everything below is a direct, server-to-server integration: no analytics platform OAuth, no third-party schema to connect.

1. Connect GitHub

Go to Settings → Integrations → GitHub and set your repo owner, repo name, and a personal access token with repo scope (or leave the token blank to fall back to the GITHUB_TOKEN environment variable). This is what lets Tamda open PRs on your behalf.

2. Tell Tamda about your product

Go to Settings → Workspace and fill in your product description, key areas, core event names, tech stack, and, if you have one, your feature flag pattern and event tracking pattern. This profile is injected into every Claude prompt, so the more specific it is, the more relevant the generated experiments and diffs are.

Tip · Already have a feature-flag or experimentation system, like Amplitude Experiment, LaunchDarkly, or a homegrown one? Paste an example under "Feature flag pattern." Tamda-generated PR diffs will read that system's existing variant assignment instead of inventing their own.

3. Send events from your app

Copy your workspace's ingest key from Settings → Integrations → Ingest key (or grab it during onboarding), and have your production app call POST /api/events whenever a user is exposed to or converts on an experiment. See "Sending events" in the sidebar for the exact request shape.

4. Start your first cycle

Go to Experiments → New. Tamda will surface the first opportunity from your event data, or you can describe a hypothesis directly. Tamda generates the spec, writes the code, and opens a PR.

5. Merge the PR and watch outcomes flow

Once the PR is merged, exposure and conversion events from your app flow into Tamda through the same POST /api/events calls wired into the diff. The Results tab shows live counts per variant, and Tamda evaluates the result once enough data has accumulated.