This field guide argues that a company should make important knowledge usable in the forms its work requires: documents and retrieval, operational records, rules, and, where relationships drive the question, structured representations such as knowledge graphs. It seemed incomplete to make that argument and then ship only a PDF. So the book is available to agents as well as people.
The book runs a Model Context Protocol server. It is read-only, needs no key, and holds no state. Point any MCP client at it:
https://lelwpzddgqhezznctjog.supabase.co/functions/v1/bn-mcp{
"mcpServers": {
"becoming-ai-native": {
"type": "http",
"url": "https://lelwpzddgqhezznctjog.supabase.co/functions/v1/bn-mcp"
}
}
}JSON-RPC 2.0 over HTTP POST, stateless. A GET on the same URL returns the server's own description. Machine-readable descriptors live at /.well-known/mcp.json and /.well-known/agent-card.json, with a plain-language summary at /llms.txt.
| Tool | What it returns |
|---|---|
search_nodes | Ranked matches across the book graph and the workshop methodology graph. |
get_node | One node in full, with its connections. |
get_book_section | A section with its callouts, figures, questions, and citations. |
get_section_prose | The actual paragraphs of a section, by number, for example 5.2. |
get_methodology_index | The session-design patterns: modules, room configurations, assignment exemplars. |
The second graph is worth knowing about. Alongside the book there is a workshop methodology layer, built from twenty years of running collaborative sessions: how to size a room, how to shape a day, what a breakout assignment contains, how work travels between rounds. An agent that reaches it can design a working session rather than only describe one.
Your agent runs its own model and pays its own inference. Nothing here proxies a model, stores a conversation, or asks for a credential. The tools are lookups over files, and there is no write path of any kind.
Very few books expose both their prose and a structured index to agents today. Building it now lets readers test whether the added interface improves navigation, comparison, and session design. The structure does not make every answer correct or create an advantage by itself; it makes the material easier to inspect and evaluate in another mode.
There is also a simpler reason. A field guide that tells leadership teams to make their institutional knowledge queryable, while remaining a stack of prose that only a human can open, would be asking for something it had not done.