Skip to main content

​
Let your agent set it up

You need Node.js 22 or newer and a coding agent that keeps its history on this machine. See agents for the ones ResumeContext reads.
1

Start your coding agent in the project folder

Claude Code, Cursor, Codex or opencode — whichever you already use. The folder you start it in is the folder that gets archived.
2

Paste this to it

It reads a page written for it and does the rest: installs the CLI, signs you in, creates the project, connects itself over MCP and checks that it all works.Keep the part in parentheses: some agents answer a URL by searching for it rather than fetching it, and curl always works.
3

Answer the two things only you can

It asks before installing anything, and it hands you a sign-in link to open in your browser. Everything else it does itself.
That link is this documentation as plain markdown, which is what an agent can read. The rendered page is setting up with an agent.

​
Check it worked

Ask your agent a question that requires the ResumeContext MCP server:
If it lists the project’s members, you are set. If it cannot tell who has access to the ResumeContext project, the MCP server is not connected — have your agent run resumecontext verify, which reports exactly which part of setup is missing, or check Connect your agent manually.
Just set up, and the archive looks empty? The first sync runs within about 20 seconds and uploads what your agents have already recorded in that folder. resumecontext sync forces it immediately.

​
Joining a project someone invited you to

Exactly the same: start your agent in your clone of the project and paste the same line. The page it reads covers the invite case, and tells it to run resumecontext accept for you.

​
Setting it up by hand

Prefer to run the commands yourself? This is what the agent does.
1

Install the CLI

Install it globally (-g). Background sync runs this installed command, so it is not set up when you run the CLI through npx. See Where to install.
2

Sign in from this machine

This opens your browser, starts the background sync service, and binds this machine, so every session it records is attributed to it. That is what lets your agent later tell you “this ran on your other laptop” or “Priya ran this, not you”. On a server with no browser, add --no-browser.
Run auth on every machine you work from. A machine that has not signed in cannot sync.
3

Create the project

Run this inside the directory the project belongs to. init creates the project and writes a marker file tying that directory to it, which is why there is no way to create a project from the website.It also starts the background daemon. From this point the directory’s agent sessions sync on their own, roughly every 20 seconds. There is no sync command to remember.In a clone of a project you have been invited to, run resumecontext accept instead — init would create a second, separate project. See Teams and access.
4

Connect your agent

Prints the MCP URL, your token, and a config block to paste into your agent. Add it only to this project folder, never to the agent’s global MCP configuration. See Connect your agent manually for the folder-level instructions for Claude Code, Cursor, Codex, and opencode.
5

Check the setup

Reports every part of setup, and pushes anything new as proof. See verify.

​
Where to next

What your agent can do

The six tools ResumeContext adds, and how to ask for them.

Syncing and freshness

What gets captured, how often, and what lags.