Open source · No signup · No API key

Any GitHub repo's docs, as a help center.

Search and browse a repository's Markdown, get answers quoted from the exact page with a link to it on GitHub, and give the same docs to your own AI over MCP.

github.com/vercel/next.js

● Free · Indexed in seconds · Nothing stored about you

Answers in your browser, and inClaudeChatGPTCursorCodex

helpdock.flisk.si/vercel/next.jsSample answer

How do I add middleware?

Quoted from the docs

1The middleware.js file convention has been deprecated in Next.js 16 and renamed to proxy.js. All functionality remains the same.

03-file-conventions/middleware.mdx ↗

2How to upgrade to version 16 › middleware to proxy

mv middleware.ts proxy.ts
upgrading/version-16.mdx#middleware-to-proxy ↗
0models called and API keys needed
1MCP server per repository, at /owner/repo/mcp
6 hat most between rebuilds from GitHub
MITopen source, and runs on your machine

See it run

From a GitHub address to a cited answer.

Swap the host, ask in your own words, follow the quote back to the file, then hand the same docs to your AI.

A 23-second film: github.com is replaced with helpdock.flisk.si in the address of vercel/next.js, its docs open as a help center, the question "How do I add middleware?" is answered with passages quoted from the docs and links to them on GitHub, and the "Use in your AI" panel shows the MCP address.

A help center

Ask in your own words. Answers are passages quoted from the docs, each linked to its file and heading on GitHub. Every page is browsable.

An MCP server for your AI

Every repository gets one at /owner/repo/mcp, for Claude, ChatGPT, Cursor or Codex:

claude mcp add --transport http nextjs https://helpdock.flisk.si/vercel/next.js/mcp

Your AI, not ours

Helpdock never calls a model and needs no key. Under each answer, one click hands the question and the cited passages to your ChatGPT or Claude.

Reads the repoThe public archive from GitHub, Markdown and MDX only, translations and tooling skipped.
Indexes it in secondsSQLite full-text search with stemming and ranking. No vector database, no embeddings.
Keeps nothing about youThe URL is the state. Indexes are rebuilt from GitHub every six hours.
Runs anywhereOpen source, built and run with Flisk. Point it at your own docs with flisk up.