Docs

Setup guide

Digg captures the AI-coding sessions you approve, strips secrets and personal data on your machine, and earns you AI Stocks for every session that enters the training corpus. This guide walks you through setup end to end.

1. Choose how to install

Digg is a terminal CLI and a background daemon. Install it yourself, or have your coding agent do it — either way you end up with the same thing.

Terminal CLI

One digg command. No editor needed. Node 18+. This is the recommended path.

Let your agent do it

Paste one line into Claude Code, Codex, or Cursor and it runs the setup, asking you before each step that needs consent.

2. Set up the terminal CLI

Requires Node 18 or newer (Node 22+ also captures Cursor's in-editor chat). Works on macOS and Linux.

  1. 1
    Install the CLI
    This downloads the digg command to a folder on your PATH.
    curl -fsSL https://digg.sh/cli.sh | bash
    If it warns that the folder is not on your PATH, copy the export PATH=… line it prints into your ~/.zshrc (or ~/.bashrc) and open a new terminal.
  2. 2
    Run guided setup
    One command signs you in, connects this machine, and starts capturing:
    digg
    Prefer to do it step by step? Run digg login, then digg connect, then digg start.
  3. 3
    Approve the sign-in in your browser
    Digg prints a link like https://app.digg.sh/link?code=… and a short confirm code. Open the link (you must be signed in to Digg in that browser), check the code matches, and approve. No account yet? Create one first.
  4. 4
    Confirm it is running
    digg status
    You should see Signed in: yes, Connected: yes, and Capturing: yes. From here, capture runs in the background and uploads on its own, with no further commands. Secrets are stripped before anything leaves your machine.
  5. 5
    Capture your past sessions (optional)
    To include the coding history you already have on disk, run:
    digg backfill
    then restart capture with digg stop and digg start. Already-uploaded events are skipped, so it is safe to run more than once.
  6. 6
    Watch it fill up
    Check anytime with digg earnings and digg sessions, or open your dashboard. Publish a session to Discover with digg share <id> public.
Reboots are handled for you. Guided setup registers the daemon with launchd (macOS) or systemd (Linux) as a user service, so capture resumes at login without you remembering anything. It never uses sudo. Turn that off with digg service uninstall, or check it with digg service status. A bare digg start on its own survives closing the terminal but not a restart.

2b. Let your coding agent set it up

You already have an agent with a terminal. Paste this into Claude Code, Codex, Cursor, or anything else that can fetch a URL and run commands:

Read https://digg.sh/start.md and set me up on Digg.

digg.sh/start.md is a setup procedure written for an agent rather than a person. It installs the CLI, prints the sign-in link for you to approve in your browser, connects the machine, and registers the background service.

It has to ask you. The procedure tells the agent never to enable capture, widen it to every repository, or upload your existing history without your explicit yes, and never to complete sign-in on your behalf. If an agent does any of those without asking, that is a bug worth reporting.

For agents that support installable skills, the same instructions ship as a reusable skill in skills/digg/ so you can add it once and then just say “check my Digg status” or “pause capture for Codex”.

3. How you earn

Per approved session

Every session that clears the quality bar and enters the training corpus earns you AI Stocks.

Track your holdings

Your dashboard shows what you earned today, over time, and your current position.

Nothing to buy

Digg is free. You contribute the sessions you approve; the coding you already do is the work.

Per-session rewards are capped during launch (a daily and a monthly ceiling) to keep things fair. Your dashboard shows how much of each window you have earned.

4. Privacy and redaction

Redaction happens on your machine, before anything is uploaded. Digg removes three classes of sensitive data and reports a counts-only summary per session, it never stores or shows the removed value itself:

When a batch reaches the server it is re-scanned as a second gate. Each session on your sessions page shows the exact counts removed and anything the server caught on re-scan. You can delete any session at any time.

5. Manage or stop capture

Pause or stop
  • digg stop stop capturing now
  • digg start resume
  • digg consent pause claude-code pause one tool
  • digg consent resume claude-code resume that tool
Remove it
  • digg service uninstall stop it resuming at login
  • digg logout sign out on this machine
  • digg uninstall remove all local Digg state
  • Sessions already contributed are managed from your dashboard

6. Command reference

digg
Guided setup: sign in, connect this machine, and start capturing, all in one.
digg login
Sign in with the device-link flow (approve in your browser).
digg logout
Sign out, disconnect this machine, and stop capturing. Clears the local enrollment so the next person enrolls under their own account.
digg switch
Sign out and set up a different account in one step.
digg connect
Connect this machine for capture (enroll). Records your consent.
digg start
Start capturing in the background. Uploads run on their own from here.
digg stop
Stop capturing.
digg service install
Resume capture automatically after a reboot (launchd on macOS, systemd on Linux).
digg service uninstall
Stop capture from resuming at login, and remove the unit.
digg service status
Show whether the background service is installed.
digg status
Show sign-in, connection, capture, and reward status.
digg earnings
Your rewards: today, lifetime, available, plus session-reward progress. Alias: digg balance.
digg sessions
List your recent captured sessions. A filled dot means it is in the training corpus.
digg share <id> public|private
Publish a session to Discover, or make it private again.
digg consent allow-all
Capture every repository on this machine.
digg consent pause <harness>
Stop capturing one tool, e.g. claude-code, codex, cursor.
digg consent resume <harness>
Resume a paused tool.
digg backfill
Re-capture your existing on-disk history from before you installed Digg.
digg uninstall
Stop capture and remove all Digg state from this machine.
digg version
Print the installed CLI version.

7. Troubleshooting

`command not found: digg` after installing the CLI+

The install directory is not on your PATH. The installer prints the exact line to add (for example export PATH="$HOME/.local/bin:$PATH"). Add it to your ~/.zshrc or ~/.bashrc, then open a new terminal.

`dig status` printed DNS records instead of running Digg+

The command is digg, with two g's. dig is the DNS lookup tool that ships with macOS and Linux, so a single-g typo runs that instead and answers with DNS records rather than an error. Run digg status.

Nothing is capturing / no sessions appear+

Run digg status. If it is not capturing, run digg start or sign in again. New sessions show up after the server curates them, which takes a short while, so give it a few minutes after you finish coding.

The sign-in link expired+

The device link is valid for a few minutes. Just run digg login again to get a fresh link.

Capture stopped after I rebooted+

A plain digg start does not survive a restart. Run digg service install once and the daemon is registered with launchd (macOS) or systemd (Linux), so capture resumes at login.

Cursor IDE chat is not being captured+

Cursor's in-editor chat lives in a database that needs Node 22 or newer to read. Claude Code and Codex capture on any Node 18+. Upgrade Node and run digg backfill to pick it up.

FAQ

What are AI Stocks?+

Your contributions earn you AI Stocks: a stake in the AI your data helps build, allocated for every approved session that enters the training corpus. Your holdings appear on your dashboard. Eligibility, allocation, and any vesting are set out in the Terms.

What exactly gets captured?+

Your AI-coding sessions: the prompts you send, the tools the agent runs, the edits it makes, and how each task turned out. Nothing outside your coding sessions is touched: no browsing, no keystrokes in other apps, no files the agent never saw.

How is my private information protected?+

Secrets (API keys, tokens, private keys), PII (emails, IPs, phone numbers, cards), and your username in file paths are stripped on your machine, before anything leaves it. When a batch lands, the server re-scans it as a second gate. Every session shows a counts-only summary of exactly what was removed. Read the privacy policy.

Can I see and delete my sessions?+

Yes. Every captured session appears on your sessions page with its redaction summary. Delete any one and it is hard-removed, the session and all its events, from our database.

Is there a VS Code or Cursor extension?+

No longer. Digg is the terminal CLI and its capture daemon, which reads the same Claude Code, Codex, and Cursor sessions with one install path to keep secure. Install it with curl -fsSL https://digg.sh/cli.sh | bash, then run digg. Existing extension installs keep working but are no longer updated.

What is Discover?+

Sessions their owners chose to make public: a feed of real AI-coding runs you can read end to end. Sharing is opt-in and per-session; your sessions stay private unless you publish them. Browse Discover.

Still stuck? support@digg.sh

Digg: Docs & setup guide