applyer

<div align="center"> <img src="src/renderer/src/assets/logo.png" width="76" height="76" alt="Applyer logo" /> # Applyer **Turn the coding agent you already use into a focused job-search copilot.** Search, compare, queue, and draft applications in one local desktop workspace. The agent handles repetition; you control every consequential action. [](https://github.com/xCirno1/applyer/actions/workflows/ci.yml) [](https://github.com/xCirno1/applyer/releases/latest) [](LICENSE) [](package.json) [](SECURITY.md) [Download latest](https://github.com/xCirno1/applyer/releases/latest) · [Security](SECURITY.md) · [Contributing](CONTRIBUTING.md) </div> <br /> https://github.com/user-attachments/assets/c18e05b0-cf45-47c5-b748-ee69cec799bb ## One workspace, from search to review Applyer pairs an embedded terminal with a live application board. Connect an [MCP](https://modelcontextprotocol.io/)-capable CLI such as [Claude Code](https://claude.com/product/claude-code) or [Codex](https://github.com/openai/codex), or connect an OpenRouter account and chat with Applyer's built-in agent instead, describe the roles you want, and let the agent research job sites, score matches, organize the pipeline, and prepare forms in a visible browser. Final submission stays with you. - **Search with context** across Indeed, LinkedIn, Seek, Jora, Prosple, and Remotive (each in the national edition for your country), plus public Greenhouse, Lever, Ashby, and Workday boards. - **Work from a real pipeline** with queued, filled, submitted, and failed states, exclusions, filters, bulk actions, notifications, and a searchable audit trail. - **Measure each run** by starting one before handing the agent a task: searches and results per site, jobs queued, forms filled, challenges hit, failures by reason and every tool call are counted live, with a timeline and a history of past runs. - **Draft multi-page applications** through inspected fields and permission-gated navigation. The agent cannot access a final-submit action. - **Keep named resume variants** ("Backend-focused", "Concise one-page") built from a structured master resume: the agent rewrites and reorders what you already have (never adds to it), each job points at the variant that fits it, you review the diff and the rendered PDF, and that is what gets attached. - **Keep your workflow yours** with movable local storage, JSON/CSV import and export, custom appearance, keyboard shortcuts, and English or Indonesian UI. ## What's new The latest release adds an in-app OpenRouter agent as an alternative to a CLI, per-run statistics, four additional job sources (Seek, Jora, Prosple, and Remotive) with country-aware search, and named resume variants with an in-app editor and PDF export. See the [changelog](CHANGELOG.md) for the concise release record. ## Security and control Trust is a product boundary, not a slogan: - **Local-first data.** Applyer has no hosted account or application-data backend. Profiles, documents, job history, screenshots, and logs stay in the storage location you choose. Network traffic goes to job sites and, depending on which agent mode is active, to the provider used by your agent CLI or to OpenRouter. - **Encryption you control.** Encrypted mode covers the SQLite database and files, with keys protected by the operating-system keychain. Plaintext mode remains an explicit choice. The recovery model is documented in [Encryption and local data](docs/encryption.md). - **Bounded automation.** The MCP bridge exposes a small, validated tool surface. Form fields are inspected before editing, browser navigation requires permission, and no tool can submit an application. Your CLI's own filesystem and shell permissions remain governed by that CLI. - **Auditable releases.** CI runs lint, type checks, tests, builds, and a high-severity dependency audit. Release artifacts are produced on native OS runners and published with `SHA256SUMS.txt`. - **Responsible disclosure.** Please report security issues privately using the process in [SECURITY.md](SECURITY.md), not in a public issue. ## Install Download the artifact for your operating system from [GitHub Releases](https://github.com/xCirno1/applyer/releases/latest): | Platform | Packages | |---|---| | Linux | AppImage, `.deb`, `.rpm` | | macOS | `.dmg`, zipped `.app` | | Windows | per-user NSIS installer, portable `.exe` | You also need an installed and authenticated MCP-capable agent CLI, or an OpenRouter account you can connect from inside the app instead. The first launch guides you through your profile, documents, storage mode, and agent connection. > **Release integrity:** Release packages are not yet code-signed or notarized, so macOS and > Windows may show an identity warning and Windows may scan the app on first launch. > Verify your download against the release's `SHA256SUMS.txt`. Never install an Applyer > binary obtained from an unofficial mirror. ### Run from source Requires Node.js 20.19+ or 22.12+ and npm. ```bash git clone https://github.com/xCirno1/applyer.git cd applyer npm install npm run dev ``` `npm install` rebuilds the native Electron modules and installs Chromium for browser automation. Development builds use a separate `applyer-dev` data directory, so they do not touch an installed copy's data. ## How the agent connects Applyer can configure Claude Code and Codex during onboarding, globally or only for Applyer's workspace. Other MCP clients can use the generated configuration. The bridge offers these capabilities: | Area | MCP tools | |---|---| | Profile | `get_profile`, `update_profile` | | Discovery | `search_jobs`, `get_job_details`, `add_company_board`, `list_company_boards` | | Pipeline | `queue_job`, `list_jobs`, `flag_failure`, `exclude_job` | | Application | `inspect_application`, `fill_application`, `edit_application`, `click_application_button` | Every profile write and application action is visible in the Activity Log. Verification challenges pause the workflow for you instead of being bypassed. ### OpenRouter mode Instead of a CLI, connect an OpenRouter account: authorization happens in your browser (no API key typed in), and the key is stored with your OS keychain. Pick any tool-capable model from OpenRouter's catalog, defaulting to `deepseek/deepseek-v4.1-flash`. The same MCP tools above run in-process, driven by a chat panel on the right of the workspace rather than a terminal: sessions you can search, rename and delete, a model switch per chat next to the input, and each tool call shown as a row that opens into its arguments and result. Data-changing tools ask for approval inline in the chat by default, configurable in Settings > Agent. Chat history is local, encrypted per your storage setting, and included in export/import along with the rest of your data. ## Development ```bash npm run typecheck # main, preload, and renderer TypeScript npm run lint # ESLint npm test # Vitest suite npm run build # production bundle npm run smoke:mcp # live MCP protocol smoke test npm run test:site # local browser-automation fixtures npm run package # package for this OS npm run package:linux # AppImage + deb + rpm npm run package:mac # dmg + zip (macOS host) npm run package:win # installer + portable exe npm run package:all -- --dry-run ``` ### Repository map ```text src/main/ Electron lifecycle, storage, browser automation, MCP, IPC src/preload/ context-isolated renderer bridge src/renderer/ React 19 interface, state, themes, and localization src/shared/ validated types, settings, and cross-process contracts scripts/ migrations, smoke tests, and packaging pipeline test/ shared mocks and local job-site fixtures ``` Contributions are welcome. Start with [CONTRIBUTING.md](CONTRIBUTING.md), follow the [Code of Conduct](CODE_OF_CONDUCT.md), and use the issue templates for reproducible bug reports or focused feature requests. ## License [MIT](LICENSE) © 2026 xCirno1. Use, study, modify, and distribute Applyer with the license notice included. The software is provided without warranty.