Repository navigation
Claude Access
Apply Desk calls Claude through the Claude Agent SDK, which ships its own Claude Code CLI. Pick one of three options.
| Option | You need | You pay | Best for |
|---|---|---|---|
| A. Subscription login | Claude Pro or Max | Nothing extra; usage counts against your plan | You already use Claude Code |
| B. Subscription token | Claude Pro or Max | Same as A | Docker, servers, no browser |
| C. API key | An Anthropic Console account | Per token | No subscription, or separate billing |
If ANTHROPIC_API_KEY is set, it is used instead of a subscription login.
- Install Claude Code:
curl -fsSL https://claude.ai/install.sh | bash - Run
claude, type/login, sign in. -
.venv/bin/python -m app selftestprintsok Claude login works.
The login lives in ~/.claude/. The Docker setup mounts it into the container.
- On a machine with a browser:
claude setup-token - Running directly:
export CLAUDE_CODE_OAUTH_TOKEN=<token>beforepython -m app serve. Docker:cp app/.env.example app/.envand setCLAUDE_CODE_OAUTH_TOKEN=<token>.
- Create a key at console.anthropic.com and add credit.
- Running directly:
export ANTHROPIC_API_KEY=sk-ant-...Docker: setANTHROPIC_API_KEY=sk-ant-...inapp/.env.
You do not need Claude Code installed for this option. You do need it once for the /setup interview, or fill the profile by hand (First Run).
app/.env is git-ignored. Never commit a key or token.
Amazon Bedrock and Google Vertex AI work through the usual Claude Code variables (for example CLAUDE_CODE_USE_BEDROCK=1 plus AWS credentials). Model IDs in Settings must then be valid for that provider.
| Step | Default model | Volume |
|---|---|---|
| Triage | Claude Haiku | many short calls (titles only) |
| Scoring, Gmail replies | Claude Sonnet | tens per day |
| Drafting, interview prep | Claude Opus | up to "Drafts per run" |
Drafting is the biggest cost. To spend less: lower Drafts per run, raise Minimum fit score, set the drafting model to Sonnet, raise Minimum company size, or turn off schedules you don't need.
Apply Desk · MIT · Built on AI Job Search · Screenshots use fictional demo data
Get started
Using it
- Dashboard Tour
- Daily Search Pipeline
- Job Sources
- Gmail Tracking
- Drafts and Interview Prep
- Schedules and Notifications
Reference
Help