> ## Documentation Index
> Fetch the complete documentation index at: https://infino.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Lantern

> Ask any public GitHub repo questions in plain English. Answers come from Infino search, with sources you can click.

Lantern is a live demo of Infino as the retrieval layer for an AI agent.
Pick a repository, ask a question in plain English, and Claude answers by
searching the code through Infino. Every answer cites the exact lines it
used and shows what it took: search in milliseconds, thinking in seconds,
tokens spent.

<video autoPlay muted loop playsInline controls className="w-full rounded-xl" src="https://mintcdn.com/infino-29/bzyG_G9KLgxDQWqu/videos/lantern-demo.mp4?fit=max&auto=format&n=bzyG_G9KLgxDQWqu&q=85&s=7c40acf755a121f38a82ed293a38edd6" data-path="videos/lantern-demo.mp4" />

<Card title="Open Lantern" icon="arrow-up-right-from-square" href="https://lantern.infino.ai" horizontal>
  Runs in the browser, no login. Each visitor gets a handful of questions a day.
</Card>

## What to try

* Ask the pre-indexed `infino-ai/infino` repo something conceptual, like
  *why does a table need an append before it can be reopened?*
* When the benchmark bar appears, open it. The same question also ran
  through plain file tools (grep and read), so you can compare tool
  calls, tokens, and time side by side.
* Open **under the hood** to see what the index actually stores: the
  schema, sample chunks, and the files Infino wrote on disk.
* Paste any public GitHub repo to index it live. Large repos start
  answering while indexing is still in progress.

## How Infino powers it

| In the demo                      | Infino feature                                                            |
| -------------------------------- | ------------------------------------------------------------------------- |
| Keyword search tool              | BM25 full-text search                                                     |
| Semantic search tool             | Vector search over the chunk embeddings                                   |
| Hybrid search tool (the default) | Fused keyword and vector ranking in one call                              |
| Under-the-hood inspector         | SQL over the same table, plus the literal storage layout                  |
| Live indexing                    | One table per repo with full-text and vector indexes, appended in batches |

The agent talks to Infino through the same tool names the
[Infino MCP server](/integrations/mcp) exposes, so what you see is what
any MCP-connected agent gets.