WikiAIG Cookbook

What WikiAIG is

A place to put structured knowledge so your AI tools can read it directly — through MCP, plain markdown URLs, /llm.txt, and /context.json.

That's the whole product. The rest of this page is what that actually means, and why it's worth your time.

The problem

You have docs that your AI tool keeps getting wrong.

You've tried pasting them into the context window — too big, blows your budget, model misses things. You've tried RAG — works okay for retrieval but breaks down when the questions need structure ("compare these three approaches", "what changes in v2 vs v1"). You've tried letting the model crawl your docs site — slow, unreliable, you have no idea what it actually read.

The real shape of the problem: agents don't need your documentation pages, they need your knowledge. Pages are an artifact of how humans read docs. The knowledge underneath — entities, concepts, comparisons, claims, sources — is what the model is actually trying to recover from the prose.

What WikiAIG does about it

You push structured wiki pages (markdown, optionally with rich HTML) through MCP from your AI tool of choice — Claude Code, Cursor, Codex CLI, Claude Desktop, anything MCP-aware. WikiAIG hosts them. Then any MCP client can read them back. Same wiki, every tool.

The wiki is the artifact your tools read. Pages are typed — overviews, concepts, entities, comparisons, sources — so the model can ask for the shape of knowledge it actually needs, not just "the next thousand tokens that look relevant."

You can also publish wikis publicly, and other people's MCP clients can ground on them. That's the platform side.

Who this is for

• Engineers maintaining internal docs that an AI tool needs to read accurately
• Tool builders who want to ship their docs as MCP grounding alongside the docs site
• Anyone publishing technical knowledge that should be readable by both humans and agents

Who this isn't for

• General-purpose note-taking — use Notion or Obsidian
• Personal scratch — local files are fine
• Replacing your docs site — WikiAIG complements it, doesn't replace it

Open vs hosted

Everything you push is yours. You can export, fork, mirror. Public wikis are crawlable and citable. We host because someone has to run the MCP server and keep the data fresh — the structure and the content are yours, and the export is one CLI command away.

Where to next

→ [Connect Claude Code / Cursor / Codex CLI](connect-your-tool) — the canonical wire-up.

→ [Pushing content via MCP](pushing-content-via-mcp) — the developer workflow with real code.

→ [HTML hosting — capabilities, constraints, examples](html-hosting) — and the visual showcase that comes with it.