Alfred
Your job-hunt butler.
Track every application you have in flight, and turn each one into work you can
actually do — a calibrated fit score, a prep plan, and the questions the loop will ask.
Documentation · Quick start · Providers · Full demo ↓
What it does
| |
|---|
| Track | Nine stages, drag and drop, and a timeline recording every move, note and email. Paste a posting and Alfred extracts the fields. |
| Analyze | A fit score calibrated like a hiring manager, with skill-by-skill coverage, the gaps worth closing, and a positioning angle. |
| Prepare | Coding problems, concepts, system-design prompts and behavioural stories chosen for that role — each with a reason. |
| Rehearse | The questions this loop will probably ask, with answers drafted in your own voice for you to rewrite. |
| Ingest | Optional read-only IMAP sync that classifies recruiter mail and matches it to your pipeline. |
It runs locally against SQLite. Nothing leaves your machine except the calls you
configure to an AI provider.
Quick start
git clone https://github.com/Unchained-Labs/alfred.git
cd alfred && npm install && npm run dev
Open http://localhost:3000. The database is created and migrated on first
render — there is no setup step. Requires Node 22+.
Want to see it populated first? npm run seed loads a demo pipeline
(destructive — it replaces existing data).
Then, in Settings: paste your résumé, pick a provider, and add an
application.
Bring your own AI
One interface, four implementations — switching is a dropdown and nothing else
changes.
| Provider | Notes |
|---|
| Claude | Anthropic API — structured outputs, adaptive thinking, configurable effort |
| Local Claude Code | The claude CLI already on your machine. No API key, uses your existing sign-in |
| LLM endpoint | Any OpenAI-compatible server: Ollama, vLLM, LM Studio, OpenRouter |
| Custom agent | Your own agent over HTTP, given a documented task envelope |
A local endpoint or the local CLI keeps your résumé, the job descriptions and
your mail entirely on your machine.
Compare the providers →
Demo
The loop above is the short version. The full 60-second walkthrough plays on the
documentation site,
which also has a clip for each section — the board,
fit analysis,
prep plans and questionnaires,
the mailbox and
providers.
Or download the MP4 (1440×900, 4.8 MB).
Documentation
Contributing
Issues and pull requests welcome — see CONTRIBUTING.md. Run
make check before pushing.
A note on secrets. API keys and your mailbox password are stored in
plaintext in data/alfred.db (gitignored). That's a deliberate trade for a
single-user local tool, and it means the file is as sensitive as what's in it.
Set ANTHROPIC_API_KEY, ALFRED_LLM_API_KEY or ALFRED_AGENT_API_KEY in the
environment to avoid storing them at all — the environment wins.
Security model →
License
MIT © 2026 Erwin Lejeune and contributors
