Docs for agents

Introduction

These docs are written for two readers: you, and the model working next to you. Everything you read on navree.com is also published in the form a model reads best, plain Markdown at a predictable URL, and every app you create starts with instructions for the agents that will work on it.

They are generated from the same source as the pages you are reading, at the same time, so a model never reads docs older than the site.

Every page as Markdown

Replace the trailing slash of a docs URL with .md and you get the page as Markdown:

The Markdown starts with the page's title as a heading and its description, then the body as written: headings, tables, and code blocks with their language and file name. The site's components become plain Markdown, so nothing is lost on the way: a callout becomes a block quote that starts with its kind ("Note", "Tip", "Warning"), each tab becomes a titled section, steps become numbered headings, link cards a list of links, a file tree a text block and a diagram one sentence that describes it. The files are served as text/markdown, with no navigation, scripts or styling around them.

Point your agent at the page for the task:

Read https://navree.com/docs/digging-deeper/queues.md, then add a queued job
that exports a company's contacts to CSV and mails the requester a link.

llms.txt

/llms.txt follows the llms.txt convention: a short description of Navree, then one line per docs page, in the order of the sidebar, linking the page's Markdown with its one-sentence description:

# Navree

> The complete backend framework for Rust. Modular, API-first and AI-native. Everything your backend needs, checked by the compiler.

Navree is a batteries-included Rust framework for building secure, scalable server-side applications. ...

## Prologue

- [Introduction](https://navree.com/docs/prologue/introduction.md): What Navree is, the ideas it is built on, what comes in the box, who it's for and how to read these docs.

## Getting started

- [Installation](https://navree.com/docs/getting-started/installation.md): Install Rust and the `navree` CLI, create an app with `navree new`, and run it with `navree dev`.
- [Your first app](https://navree.com/docs/getting-started/your-first-app.md): Build a small task-list API with an entity, a migration, validation, JWT authentication, access rules and feature tests, from `navree new` to green tests.

## The basics

- [Routing and controllers](https://navree.com/docs/basics/routing.md): Controllers and route attributes, the API prefix and versions, absolute `!` paths, path and query parameters, route names and the route table.
- [Validation](https://navree.com/docs/basics/validation.md): Validate request bodies and query strings with `Valid<Json<T>>` on garde, with field-level messages and `unique` / `exists` database rules.
...

The index ends with an "Agent skills" section: one line per agent skill, linking its SKILL.md on navree.com, and the address of /skills/index.json, which lists them all.

An agent that starts from llms.txt reads the index (a few thousand tokens), picks the pages that matter for the task and fetches only those. That is the best default: small, current and specific.

llms-full.txt

/llms-full.txt is every page's Markdown in one file, in sidebar order, each page starting with its title and URL. Use it when you want the model to hold the whole framework at once: loading it into a project's knowledge in a chat app, or asking broad questions ("Which modules does an app with workspaces, queues and real time need, and in what order?").

It is large, the size of a book. For a coding task in your editor, the index and a few pages are faster and cost less; for a long design conversation, the full file saves the model from fetching page after page.

Page actions

Every docs page has actions next to its title:

"Open in" starts the conversation with a prompt that names the page's Markdown URL and asks the assistant to read it first, so your first message can be the question itself: "How do I make this job unique per company?"

AGENTS.md and CLAUDE.md

navree new writes two files at the root of every app, next to README.md:

AGENTS.md is the project's instructions for coding agents, in the format most agents read on their own. It is written for the app it lives in, with the app's name, and covers:

• Layout: what lives in config/, src/app.rs, src/abilities.rs, src/modules/<name>/, tests/ and storage/ (and web/ for apps created with --ui next).
• Commands: navree dev, cargo test, cargo clippy --all-targets -- -D warnings, cargo fmt, navree check, navree gate, navree routes, navree migrate and navree list.
• Generators: which navree make:* command to run for each kind of code, and the marker comments (// navree:modules, // navree:wire, // navree:policies) they register at.
• Conventions: deny-by-default routes and the access attributes, modules and injected services, validated DTOs, the error shape and AppError helpers, the unit of work and Repo, migrations, configuration and secrets, and tests with TestApp.
• Before you finish: format, lint and test, and fix everything they report.
• Documentation and skills: these docs, llms.txt, the Markdown pages and how to install the agent skills.

CLAUDE.md is for Claude Code, which reads that file name. It imports AGENTS.md instead of repeating it, so every agent reads the same instructions:

@AGENTS.md

The conventions for coding agents in crm live in AGENTS.md, imported above; keep them there so every agent reads the same file.

Making them yours

The generated AGENTS.md describes Navree's conventions. Add your team's: the domain words agents should use, the modules that are off limits, how you name migrations, which Navigator tools to use before changing routes or the schema. Keep it short and specific; agents read it at the start of every session, and a rule written down there beats the same rule explained again in every prompt.

Edit AGENTS.md, not CLAUDE.md. Claude Code follows the import, and every other agent reads AGENTS.md directly.

For an app created before these files existed, run navree new for a scratch app of the same kind and copy its AGENTS.md and CLAUDE.md into your project, changing the name.

Choosing what to give your agent

Skills, AGENTS.md and Navigator work together: AGENTS.md says how this app is built, a skill says how to do this kind of task, Navigator says what the app looks like right now, and the docs say everything else.