Community

Seekter Setup, Features, and Privacy Overview

Seekter's initialization involves cloning the repository, running setup commands, and configuring job-search preferences through 40 questions, with options to manage resumes and verify settings.

AI-generated editorial illustration of the article topic; not a photograph, product screenshot or measured result.

Initialization and Setup Process

The initialization and setup process for Seekter begins with cloning the repository and opening it in Claude Code. Users then run the /seekter-init command, which guides them through approximately 40 short questions to configure the job-search agent. This step-by-step process helps tailor Seekter to the user's preferences and requirements. If users need to add or replace a resume, they can drop the PDF into the designated folder and update the table in profile/profile.md to specify which CV is the default and which one corresponds to specific role types. To verify the setup, the command python3 scripts/seekter.py settings --check displays which values are still using defaults and highlights any missing required information.

Tracking and Reporting Features

Seekter’s tracking and reporting features provide detailed insights into job-search activity without compromising user privacy. The system logs every posting it interacts with, organizing them by the month they were first handled under the applications/ directory. Each month’s activity is summarized in a README.md file, listing applications with their current status. Individual job postings are stored as markdown files with timestamps and details, such as 2026-09-22--ruby-labs--senior-backend-engineer.md. The report aggregates data like funnel and response rates by source and location, highlights top skip reasons, and suggests up to two changes. Notably, the report reads from applications/ and runs/, not from email, meaning replies not logged by Seekter do not appear in the report. This design ensures transparency while respecting user control over data.

Importing and Managing Job Applications

Seekter, an open-source job-search agent running in Claude Code, allows users to import and manage job applications from various sources. To begin, users can export their existing job tracker from tools like Notion, Airtable, or Google Sheets as a CSV file. Using the command-line tool scripts/import_csv.py, they can then import this data into Seekter. The tool matches column names loosely, supporting fields such as Position/Role/Title, Company, Status, Job URL, Applied on, Source, and Notes. For best results, users are encouraged to include URLs in their CSV, as rows without URLs can only be matched by company name later. This approach ensures that duplicate job listings—whether found through LinkedIn, an aggregator, or a company site—are correctly identified and consolidated under a single, normalised job key, such as a LinkedIn ID or ATS UUID.

Repository Structure and Additional Tools

The Seekter repository is organized to support its core functionality as an open-source job-search agent. Key directories include .claude/skills/, which contains scripts like seekter-init, seekter-run, and seekter-report, and reference/, which holds job source files and application form system definitions. The scripts/ directory includes seekter.py and other utilities for managing job searches. Testing is handled via tests/, where CLI tests are run using python3 -m unittest discover tests. Additional tools and workflows are managed in .github/, including leak scans and issue templates. User-specific data, such as applications and runs, are stored in the profile/ directory and are git-ignored for privacy. Notably, there is no version file included in the repository.

Privacy and External Integrations

Seekter’s approach to privacy is outlined in its PRIVACY.md document, which details what data Claude, employers, and job sources can observe during its operation. The tool includes safeguards to minimize exposure, such as its linkedin.mode setting, which, when switched to read, allows Seekter to access LinkedIn job details and search results in a read-only capacity. This mode enforces daily request limits, includes pauses between requests, and automatically disables itself upon encountering warning or unusual-activity pages, as noted in the project’s documentation.

Additionally, the seekter-webui project by Ege BULUT provides a browser-based interface for running Seekter’s daily loop, replacing the terminal-based CLI with a standard-library server. This interface maintains the same core functionality—tracking, filtering, and reporting—but presents it through a user-friendly web interface.