JobSentinel

JobSentinel is a free, open-source, local-first desktop assistant for finding real, relevant, fairly compensated work while keeping job-search data under the user's control.

JobSentinel is an open-source, local-first job-search assistant for finding real, relevant, fairly compensated work while keeping sensitive job-search data under user control.

Find better roles, spot weak postings, tailor truthful resumes, track every application, and protect your pay goals from one private desktop workspace. Core workflows work locally. JobSentinel is free, will always stay free, and will always remain MIT licensed.

Start here: Download | Capabilities | Quick start | Privacy | Research | Contribute

Local dashboard with job matches, posting-risk cues, filters, source status, and dark mode.

What It Helps With

JobSentinel is for job seekers, not hiring teams. It is designed for technical and non-technical roles, career changes, long searches, and people who should not need terminal commands or debugging skill to get value.

The Short Version

Flow	What JobSentinel makes easier
Find	Search approved sources, employer career pages, regional boards, and user-opened job sites from one local workspace.
Judge	See fit, freshness, source trust, pay transparency, reposting clues, and likely ghost-job signals before spending time.
Prepare	Match resumes to real requirements, catch hard gaps, improve readability, and draft only claims the user can support.
Apply	Review forms, screening questions, attachments, notes, reminders, and follow-ups without letting the tool submit for you.
Negotiate	Keep salary floors, written offers, verbal numbers, total compensation, commute, relocation, and deadlines in view.

Need	JobSentinel response
Find jobs without losing control	Check approved sources, open search links, import visible jobs, and keep a local job board.
Avoid stale or low-trust postings	Review source, freshness, reposting, pay, and ghost-job signals before spending time tailoring.
Track applications cleanly	Save jobs, notes, contacts, reminders, interviews, offers, and no-response follow-up.
Improve resumes truthfully	Build resumes, review readable text, compare against jobs, and draft evidence-bounded improvements.
Protect pay goals	Set salary floors, compare listed pay, review written offers, and prepare local counter or decline notes.
Keep data private	Store search data locally by default, with no telemetry and no hosted account requirement.

What Makes It Different

• Ghost-job review is built in. JobSentinel looks for stale, reposted, weak-source, scam-like, and low-confidence signals so users can decide where their effort is worth spending.
• Source discovery understands the messy job web. It can classify public feeds, employer career pages, regional boards, ATS-backed postings, user-opened search destinations, and restricted sign-in sources.
• Resume help stays evidence-bound. It can help a user clarify real experience, improve readability, and map resume evidence to job requirements without fabricating qualifications.
• Restricted-source work is user-driven. LinkedIn-style workflows are built around user-opened sessions, visible-job import, local review, clear warnings, and no session-token storage.
• Pay protection is a core workflow. Salary floors, listed ranges, written-vs-verbal offers, total compensation, commute costs, relocation, and pressure deadlines are treated as first-class job-search data.
• Local match intelligence goes beyond keywords. Embedded builds can use governed Qwen3-Embedding-0.6B retrieval plus Qwen3-Reranker-0.6B reranking, combined with exact skill, BM25, seniority, blocker, and evidence signals.
• Security is part of the product mechanics. Safe support reports, local encrypted secrets, optional AI request previews, and release artifact checksums are built into normal flows instead of bolted on later.
• Agent Skills ship with the app. The release includes downloadable job hunting and resume skills for agents that support skills packages.
• External AI is optional, not the product. Users can configure providers, preview payloads, approve requests, and keep the local path available.

Download

Use the latest GitHub release and verify the matching .sha256 checksum from the same release before opening a package or installer.

Platform	What to use
Windows 11+	Use the .msi or setup .exe. If the filename includes _unsigned, expect SmartScreen and verify the checksum first.
macOS	Use the _no-account_universal.dmg for Apple silicon and Intel Macs. Until Apple Developer Program materials are available, the package is not Developer ID signed and not notarized, and may require first-open Privacy & Security approval.
Linux	Use the .AppImage or .deb when present for the release.
Agent Skills	Download the Agent Skills ZIP on Windows, or ZIP/tar.gz on macOS and Linux, then verify the matching checksum.

JobSentinel does not silently update itself in 2.9.0. Back up Settings before replacing the app, then install the newer release. See Updating Or Going Back.

Agent Skills downloads are separate from the desktop installer. Release assets include JobSentinel-X.Y.Z-agent-skills.tar.gz and JobSentinel-X.Y.Z-agent-skills.zip with matching .sha256 checksums. Use the ZIP archive on Windows. It is the most portable extraction path.

First-open prompts

macOS can show:

JobSentinel can't be opened because Apple cannot check it for malicious software.

Continue only if you downloaded JobSentinel from the release page above, expected this file, and verified the .dmg with the matching .dmg.sha256 from the same release. If you cannot verify it, delete the download and wait for a replacement package or build locally.

Windows SmartScreen can show:

Windows protected your PC

Continue only if the file came from the release page above and the checksum matches. Unsigned Windows files must be clearly labeled _unsigned.

Why Trust It

Rule 0: user privacy and security are non-negotiable. No feature, integration, support shortcut, external AI provider, source adapter, or test fixture may weaken local control, credential safety, explicit review, or privacy-preserving defaults.

Boundary	Commitment
Local-first data	Searches, saved jobs, notes, resumes, salary floors, applications, and support reports stay local by default.
No telemetry	JobSentinel does not collect analytics or behavioral telemetry.
External AI	Optional, disabled by default, preview-gated, and routed through the privacy-first AI gateway.
Secrets	Alert credentials and source access codes use the local encrypted vault and OS credential store where available.
Applications	JobSentinel helps prepare and track applications, but the user reviews and submits outside JobSentinel.
Restricted sources	User-gated, warning-backed, local, and review-first. JobSentinel does not store restricted-site cookies, tokens, browser storage, or auth headers.

Current macOS full-public-readiness: 94%; no-account path completion: 100%. A zero-friction Mac release still requires Apple Developer Program materials, Developer ID signing, notarization, stapling, and Gatekeeper acceptance.

External AI, including OpenAI or another provider, is optional, disabled by default, and routed through the privacy-first AI gateway. External requests send only needed details, show those details before anything leaves the device, and provide redaction, cancellation, approval, and local request logging.

Read more in PRIVACY.md, RESPONSIBLE_AI.md, and the security docs.

What Is New In 2.9.0

• Browser Import review queue for visible jobs the user chooses to send into JobSentinel before saving.
• LinkedIn-compatible Workbench model for user-opened sessions, local ledger actions, visible-card import, and explicit restricted-source acknowledgement without cookie, token, storage, auth-header, or hidden-page capture.
• Broader job-source taxonomy for public APIs, public employer pages, regional boards, restricted public boards, authenticated sources, and no-account limits.
• Resume Builder, Resume Match, hard-requirement review, ATS/readability guidance, evidence-bounded bullet drafting, and JSON Resume import/export.
• Local semantic matching architecture with governed Qwen3 embeddings, bounded Qwen3 reranking, hybrid skill/BM25 scoring, deterministic fallback, and privacy-safe diagnostics.
• Application Assist for review-first form guidance, screening-answer review, cover-letter checks, and manual final submission.
• Pay Protection for listed-pay gaps, written-vs-verbal offers, commute/relocation review, deadline pressure, and local counter/decline notes.
• Downloadable Agent Skills for job-search planning, posting-risk review, resume tailoring, application review, tracking, outreach, interview prep, and offer/pay review.
• Optional external AI provider configuration for OpenAI, Anthropic, Google Gemini, GitHub Copilot, and custom HTTPS providers through the local gateway.

Full notes: docs/releases/v2.9.0.md.

Job Source Terms And User Control

JobSentinel does not encourage, endorse, or hide violations of any job site's terms. The safe default path is official feeds, public employer career pages, user-opened search links, manual entry, and single-job imports the user reviews before saving.

Some sites say third-party software that scrapes, modifies, or automates activity can violate their terms, lead to account restrictions, or raise legal and privacy concerns. JobSentinel makes that risk clear before restricted-source actions. Users must explicitly accept the warning before those actions continue.

Restricted-source sessions stay user-driven:

• No hidden background access.
• No stored login details, cookies, browser storage, tokens, or auth headers.
• No CAPTCHA solving or bypass of platform controls.
• No resale, republication, or bulk redistribution of job-board data.
• No automatic final application submission.

Public unauthenticated sources can be lower-friction when their access pattern allows it, but they still use respectful request pacing, safe errors, and local review. See Job Sources.

Core Capabilities

Area	Capability
Job discovery	Approved public source checks, public employer pages, user-opened search links, Browser Import, and manual entries.
Source intelligence	Source status, duplicate cleanup, public API and ATS classification, and restricted-source boundaries.
Job review	Fit, freshness, pay transparency, ghost-job cues, source risk, and next actions.
Resumes	Resume library, builder, PDF/DOCX/JSON Resume export, ATS/readability review, and match review.
Applications	Application board, notes, contacts, reminders, interviews, offers, no-response review, and Application Assist.
Pay	Salary floors, listed-pay review, total compensation prompts, written-offer separation, and negotiation notes.
Local matching	Optional Qwen3 embedding plus reranker pipeline with exact skill, BM25, blocker, seniority, and evidence signals.
Alerts	Desktop notifications plus optional email, Slack, Discord, Teams, and Telegram after setup.
AI	Optional provider gateway with preview, approval, redaction, cancellation, and metadata-only local logs.
Skills	Downloadable Agent Skills packages for job hunting and resume workflows.

For the full maintained list, use Features And Capabilities.

Under The Hood

JobSentinel is more than a spreadsheet with job cards. The release includes several local-first mechanics that make the app useful without turning the user into a system administrator.

Mechanic	Why it matters
Local Qwen3 match intelligence	Compares jobs and resumes with semantic retrieval, reranking, exact skills, blockers, and evidence instead of only keyword counts.
Source taxonomy and routing	Separates official feeds, employer career pages, ATS families, regional boards, public boards, restricted sources, and manual paths.
Restricted-source Workbench	Lets users work from sign-in-backed sites through visible, user-started sessions without storing cookies, tokens, browser storage, or hidden page state.
Evidence-bounded resume help	Connects job requirements to real resume passages, readability issues, hard gaps, and truthful draft improvements.
Pay and offer protection	Keeps salary floors, listed ranges, written offers, verbal numbers, commute, relocation, and deadline pressure in one review flow.
Privacy-first AI gateway	Lets users configure providers, preview payloads, redact or cancel, approve sends, and keep metadata-only local logs.
Local vault and safe reports	Stores saved secrets locally and creates sanitized support reports the user can review before sharing.
Downloadable Agent Skills	Packages job-search workflows as portable skills for compatible agent tools, with both ZIP and tar.gz release assets.
Release supply-chain checks	Publishes checksums, SBOMs, attestations, no-account labels, unsigned labels, and public asset verification.

Local Match Intelligence

JobSentinel's strongest resume/job matching path is local and inspectable. When an embedded-ML build has verified model files, matching uses:

Layer	What it does
Qwen3-Embedding-0.6B	Retrieves resume and job passages that are meaningfully related, not only exact keyword matches.
Exact skill and BM25 scoring	Keeps concrete tools, certifications, titles, and requirement words grounded.
Qwen3-Reranker-0.6B	Reranks a bounded top-K set so direct evidence beats keyword-only near misses.
Blockers and evidence classes	Flags missing hard requirements, salary/location mismatch, seniority mismatch, and weak evidence.
Model governance	models.lock.toml pins model identity, revisions, hashes, sizes, licenses, and instruction profiles.

If local model files are not present, JobSentinel falls back to deterministic matching. Resume and job text stays local either way. See Local Semantic Matching and the semantic matching research notes.

Screenshots

Track applications from saved to offer, with notes and follow-up reminders.

More screenshots

Resume builder	Resume matcher

Hiring-system transparency	Application Assist

Pay Protection	Hiring Trends

Settings

Architecture

JobSentinel is a desktop app, not a hosted job-search service.

React 19 + TypeScript UI
  -> Tauri command boundary
  -> Rust 2021 services
  -> local SQLite database
  -> approved source checks and optional channels

Boundary	Rule
Renderer to backend	UI calls typed Tauri commands and handles user-safe errors.
Storage	SQLite stores searches, saved jobs, notes, applications, resumes, reports, and local request logs.
Job-source checks	Approved sources only, with respectful pacing and clear errors.
External alerts	Work only after the user configures them.
External AI	Must go through the privacy-first AI gateway. Scattered provider calls are not allowed.
Research mode	Public postings and synthetic candidate profiles are the default evaluation data.

Developer entrypoints:

• Developer architecture
• Getting started
• Testing
• Releasing
• Harness engineering

Build From Source

Requirements:

• Node.js 24.17.0 from .nvmrc
• npm 11.17.0 from package.json
• Rust 1.96.0 from rust-toolchain.toml
• Platform packages from developer setup

Current backend surface: 212 registered Tauri commands.

git clone https://github.com/cboyd0319/JobSentinel
cd JobSentinel
node scripts/install-pinned-npm.mjs
npm ci --ignore-scripts
npm run tauri:build

Common verification:

npm run harness:check
npm run lint:bloat
npm run lint:docs
npm run lint
npm run test:run
npm run test:e2e:smoke
npm run build
cd src-tauri && cargo fmt --all -- --check
cd src-tauri && cargo clippy -- -D warnings
cd src-tauri && cargo test --lib

Release verification is broader. Use docs/harness/verification-matrix.md.

Support

The easiest support path starts inside JobSentinel:

1. Open Settings.
2. Choose Save Safe Support Report or Copy Safe Support Report.
3. Review the report.
4. Share it only if you want help.

GitHub is optional. If you already use it, open the help page or start a discussion.

Project Map

Need	Link
First install and setup	Quick Start
Full capability list	Features And Capabilities
Update or roll back	Updating Or Going Back
Privacy and data boundaries	PRIVACY.md
Responsible AI	RESPONSIBLE_AI.md
Research notes	docs/research
Roadmap	ROADMAP.md
Changelog	CHANGELOG.md
External references	docs/references.md

References and external sources

The complete reference index lives in docs/references.md.

Open source. Local-first. Built for job seekers who value privacy.

Download