RepoLanternMake your first guide
Sign in, three steps to set up, then approve the estimate. Nothing is spent until you approve.
1. Sign in
Open the app and sign in with Google. Verify your e-mail address if asked.
Settings (top right) has Appearance (light, dark or follow your system), Two-step sign-in (press Add an authenticator app, scan the code with Google Authenticator, Microsoft Authenticator, 1Password or similar, and type the 6-digit code: recommended for everyone, required for admins), connecting an AI assistant (details) and your data (details).
2. Repository
- Paste the repository's
https://link. Any public HTTPS git host works: GitHub, GitLab, Bitbucket or your own. - Private repository? Tick It's private and paste a read-only access token (how to make one). It is used once to fetch the code and never stored.
- Press Continue.
3. Model
- Choose the Provider and type or pick the Model.
- Fill in what the provider asks for (an API key, an endpoint, a region…). Each provider links to its setup steps. Still needed under the buttons lists anything missing.
- Press Test connection to check before anything is spent.
- The price line shows where the model's price comes from (Costs).
Advanced (all optional):
| Option | What it does |
|---|---|
| Your price | Your own input and output price per million tokens, for a model without a listed price or a contract price |
| Audiences | Beginner, senior engineer and architect guides (up to three, from one analysis) |
| Thinking and temperature | How much the model thinks while reading the code and while writing; temperature where the model accepts it. Choices the model doesn't support are explained before you start |
| Your model's limits | Requests and tokens per minute and per day, from a dated preset or your own numbers (details) |
| If the model runs out | Up to three fallback models, each with its own keys and cap (details) |
| Length | Concise, standard or detailed sections |
| Spending limit | A hard stop for the whole guide, from $0.01 to $50 |
4. Review
The review step shows exactly what will happen: the repository, the audiences, your limit, how the model will think, where your code goes (your browser → RepoLantern, which never stores your key → your provider), and the sections you will get.
- Versions as of today: RepoLantern checks the current version of every dependency in public package registries (names and versions only, never code). For a private repository, tick the box if you allow that.
- Web search for release notes (optional): switch it on and paste your Tavily key (setup), then press Test.
- Consent: tick I agree to send this repository's text to…. It names your provider and any fallback models. It is never pre-ticked.
Press Read the code and estimate. Until you tick the consent box the button reads Tick the box to continue; if you switched web search on, test the search key first (or switch search off).
5. Approve the estimate
RepoLantern reads the code (secrets are removed first) and shows: how many files and sections, the likely, low and high cost, and the expected time. Adjust the spending limit if you want, then press Generate. Or press Not now: nothing has been spent.
6. Watch it work
The page shows the stage, the files being read, the money spent against your limit, and How it works, live: the pipeline graph and every model and tool call as it happens. You can close the tab: the guide keeps being written, and it will be here when you come back. Pause stops after the calls in flight finish; Cancel stops for good.
When it's ready, see Read and use your guide.