# Keel

> Tired of looking for a job? Let us do it for you — and we'll never lie on your behalf.
>
> Keel is an open-core job-application autopilot with a truthfulness contract: it only ever claims what you tell it is true. Free and open source (Apache-2.0). 209 verified submissions. Zero lies.

Repo: https://github.com/KeelDev-tech/keel
Docs site: https://keeldev-tech.github.io/keel/
License: Apache-2.0 (see LICENSE)

## What it is

Keel is the public half of a real production job-application pipeline:
discovery → scoring → materials → verification → launch packets, with honest
automation as the product. The public loop stops at the launch packet — a
verified, prescreened bundle (form values, banded rules, hard gates, the
per-field verification protocol) plus a documented EXECUTOR CONTRACT for
whatever submission layer you attach. Keel never submits an application.

## Open-core model

Public (Apache-2.0): discovery and scoring, truthful resume tailoring, the
canonical answer bank, prescreen gates, ATS detection + capability radar,
the launch-packet builder, append-only telemetry, outcome analytics, and
the dashboard. Private by design: submission-behavior techniques ATS
vendors could fingerprint — publishing them would get everyone's pipeline
blocked (see SPLIT.md).

## Stdlib-only

Python 3.10+, zero third-party dependencies, nothing to install. The full
test suite runs from a clean checkout with `python3 -m unittest discover -s
tests`; CI runs it on 3.10 / 3.11 / 3.12.

## The truthfulness contract

1. Truthfulness gates — hard requirements the profile can't support are
   reported as gaps, never bridged with fiction.
2. Explicit confirmation — a submission counts only on explicit
   confirmation evidence. Nothing else.
3. Fail closed — unverifiable postings, unmappable required questions,
   missing attestations: park, never proceed.

## Key directories

- engines/ — all pipeline modules (flat package): scoring, resume tailoring,
  answer bank, prescreen, ATS detection, apply loop, telemetry, dashboard
- tests/ — acceptance tests (stdlib unittest, no dependencies)
- docs/ — architecture, personalization, contributing; docs/geo/ holds
  citation-ready Q&A (FAQ, honest-automation explainer, comparison,
  alternatives, stats) plus a machine-readable stats snapshot (stats.json)
- docs/assets/ — logo, wordmark, social preview, dashboard screenshot
- sample_data/ — sanitized examples (never real applications)
- site/ — the public docs site source (serves this file with JSON-LD)

## Where to start

From a clone (`git clone https://github.com/KeelDev-tech/keel && cd keel`):

```bash
./setup.sh          # "Make it mine" — personalizes your working copy
# edit data/applicant_profile.json and data/answer_bank.json with YOUR truth
./start.sh          # status overview
python3 -m unittest discover -s tests   # run the test suite
```

Start with docs/PERSONALIZE.md. The full reference is llms-full.txt
(same directory). Citation-ready Q&A lives in docs/geo/.

## The proof figure

209 verified submissions (recounted 2026-09-22; 202 with quoted
confirmation evidence, 3 with pointer evidence, 1 URL-only, 3 unevidenced —
counted, never estimated). The figure comes from the private production
pipeline's ledger; public readers cannot independently rerun the private
pipeline, and the public repo makes no submission claims of its own. Keel
claims no user counts, growth rates, revenue, funding status, or
time-windowed figures.