Docs
Getting started
-
Your first ten minutes
From signing in to an agent that already knows your project: connect, point it at the code, watch it land.
-
Your first week
One thing a day until the loop runs by itself.
-
How to connect: which agent, which path
Chat agents in a minute, Claude Code for the full path, Codex in between; what each gives you and where the tunnel fits.
-
Start a project with your AI
The one-time instruction, then "set this up in Naetive" or "map this project"; what the agent creates and how it should behave.
-
What to ask your agent
The prompts that walk the loop, from orienting to committing.
-
Reading Naetive without reading code
The daily routine for someone who runs the project and never reads a diff.
-
Starting a chat that already knows the project
What to say for a new project, an existing one, or a normal chat; the two lines that make Claude Code orient itself.
-
FAQ
Short answers to the questions new people ask most.
Connect your AI
-
Connecting Claude Code
The one-click install: approve this computer, the key never shown; what it sets up, Windows, opt-out, what can go wrong.
-
Connecting ChatGPT
Add Naetive as a connector behind Developer Mode; give it your files with the tunnel.
-
Connecting Claude on the web
Add Naetive as a custom connector in claude.ai; one URL, no install.
-
Connecting Claude Desktop
Add Naetive as a custom connector in the desktop app.
-
Connecting Codex CLI
The config block Codex needs, and how it reaches your files.
-
Installing and starting the tunnel
Start it in the project folder, or name the folder and project; several projects; what happens on a deploy; if it does not connect.
Understanding
-
Understanding: what the project knows
The word at the top, Needs you, the map of topics and entries, the folder card, and how agents read it.
-
How understanding gets recorded
Understanding commits, the two readers, observed or proposed, supersede never contradict, when commits happen.
-
Grounding: why you can trust an entry
Cited or not yet checked, re-checks and drift, and the three choices when something needs you.
-
Keeping the understanding honest
The marks, the rules agents follow, and the small part that is yours.
-
The folder: carry the understanding in your repo
Get it once, then it follows the understanding; what is inside; what the stamp means; honest limits.
Work and agents
-
Work: what your agents are doing
The bands from Who's here to Completed, what a job is, and how to read the page without reading a chat.
-
Good job updates
What an update is and is not, when to post, the three layers, a worked example.
-
A window is an agent
One window, one row; names; resuming; why it matters.
-
Several agents on one project
One job per window, one shared understanding, the scratchpad in between.
-
The roadmap: Now, Next, Later
The project plan agents keep current, and how it differs from a session roadmap.
The pages
-
Status: the front door
The one word, the four cells, and the routine.
-
Chats: your agents, running on your machine, opened here
What the page shows, how a chat keeps running, and picking up a saved one.
-
Foundations: the documents a project rests on
The longer writing the entries cite, by topic.
-
The console and the naetive command
The small program behind the Chats page, the export command, and where things live.
Concepts and troubleshooting
-
The tunnel: what it does for you
What a chat agent gets, how the relay works, and what keeps you safe.
-
What leaves your machine
Exactly what each hook, the tunnel and the console send, and how to switch each off.
-
Activity from Claude Code
What the hook reports, where it is on by default, and how to switch it off per folder or altogether.
-
Activity troubleshooting
A window missing from Who's here, two rows for one window, activity that stopped, the folder not refreshing.
-
Troubleshooting: something is not working
Start on Connect; the common cases and their fixes.