Skip to content

login.md

Latest commit

 

History

History
109 lines (82 loc) · 4.02 KB

login.md

File metadata and controls

109 lines (82 loc) · 4.02 KB
Raw

login

crabbox login authenticates the CLI against a coordinator, stores the resulting credentials in your user config, and verifies them with GET /v1/whoami. The coordinator may run on Cloudflare or Node/PostgreSQL.

There is no built-in hosted broker. You must supply a broker URL with --url, or have one already configured (crabbox config set-broker). Without a broker, the command exits with an error.

Usage

crabbox login --url <broker-url>

This starts a GitHub OAuth login: Crabbox opens the login URL in your browser and listens on a random 127.0.0.1 port for a one-use browser confirmation. The broker releases the user-scoped bearer token only when the initiating terminal presents both that confirmation and its private polling secret.

If the browser cannot open automatically, print the URL and open it manually in a browser on the same device. Cross-device completion intentionally fails closed because the browser must return to the initiating CLI's loopback listener:

crabbox login --url https://broker.example.com --no-browser

On success the command prints the resolved broker, default provider, GitHub identity, and the config path it wrote:

logged in broker=https://broker.example.com provider=aws user=alice@example.com org=example-org config=/Users/alice/.config/crabbox/config.yaml

Flags

--url <url>                              broker URL (falls back to the configured broker)
--provider aws|azure|daytona|gcp|hetzner default managed provider to store with the broker
--no-browser                             print the same-device GitHub login URL instead of opening it
--token-stdin                            read a broker token from stdin (operator automation)
--json                                   print machine-readable JSON

When --url is omitted, the configured broker URL is used; --provider likewise falls back to the configured default provider when unset.

Operator automation with --token-stdin

Trusted operator automation can skip the browser flow and write a shared broker token instead. The token is read from stdin so it never lands in shell history or the process argument list:

printf '%s' "$CRABBOX_COORDINATOR_TOKEN" | crabbox login \
  --url https://broker.example.com \
  --provider aws \
  --token-stdin

This path stores the shared operator token verbatim and should stay limited to trusted maintainers. Prefer interactive GitHub login, which issues a per-user token, for everyday use.

What gets stored

Login writes the broker URL and token to your user config (and, when --provider is set, the default provider). Inspect the result without exposing secrets:

crabbox config show
crabbox config path

Self-hosted brokers

Each broker owns its own GitHub OAuth credentials and admission policy. A separate organization or private deployment needs its own coordinator runtime, secret injection, and GitHub OAuth app.

The coordinator requires current CLIs to provide an ephemeral HTTP loopback callback using the literal 127.0.0.1 address, a random port, and a random path. Old clients that can only poll for a completed browser token are rejected before OAuth starts.

Configure that GitHub OAuth app with a callback URL that exactly matches the broker's public origin:

https://<your-broker-host>/v1/auth/github/callback

Set the same public origin in CRABBOX_PUBLIC_URL on the coordinator, then deploy CRABBOX_GITHUB_CLIENT_ID, CRABBOX_GITHUB_CLIENT_SECRET, CRABBOX_SESSION_SECRET (distinct from CRABBOX_SHARED_TOKEN), and the relevant CRABBOX_GITHUB_ALLOWED_ORG(S) or CRABBOX_GITHUB_ALLOWED_TEAMS values. A GitHub Invalid redirect_uri error means the callback URL generated during crabbox login does not match one configured on that OAuth app.

Related