Why Job Seek

Job Seek is for people who already have a sense of where they want to work. Instead of starting with reposted listings, it monitors employer career sites and their applicant-tracking systems, then normalizes each source into one searchable model.

What you get on jseek.co

	Explore. Search postings across every tracked company and combine filters without creating an account.
	Company pages. See active and last-year posting counts, filter within one company, explore similar employers, and open every role at its source.

With a free account, you can create one watchlist and use the built-in application tracker to move roles through saved → applied → interviewing → offered/rejected, record interview rounds, and review pipeline statistics.

Pro is coming soon. The planned $10/month tier adds unlimited watchlists and email alerts for new matches.

Built by Colophon Group, a small team in Switzerland — so German, French, and Italian are first-class product languages, not afterthoughts.

Use Job Seek from AI tools and code

The hosted MCP endpoint exposes read-only tools for job search, posting details, companies, taxonomies, public watchlists, and prefilled watchlist links:

https://jseek.co/mcp

It uses Streamable HTTP and does not require authentication. Add the URL as a custom MCP connector, or run the published package locally:

npx @jseek/mcp-server

See packages/mcp-server/README.md for client-specific setup and tool examples. For direct HTTP integrations, the public REST contract is available at /api/openapi.json.

Add a company

Open issues labelled company-request are companies waiting to be added. The production backlog is processed by an isolated, Hetzner-hosted Codex runner; contributors can resolve an issue with any capable coding agent that follows the repository instructions.

ws is an agent utility. It renders the workflow, manages isolated state, and enforces the validation gates; it is not intended as a hand-configured interactive wizard.

The environment needs git, an authenticated gh CLI, Python 3.13+, and web access. Install the workflow package:

pip install jobseek-crawler-setup

Then give your coding agent this task:

Run ws task --issue <NUMBER> and follow the printed instructions.

ws fetches the issue, checks for duplicates, researches the company and all relevant career boards, guides monitor and scraper selection, validates extracted data and brand assets, and opens the pull request. The registered crawler types cover common ATS APIs, sitemaps, structured data, rendered pages, PDFs, and vendor-specific formats.

No issue for the company you want? Request it. Anyone can.

The maintained workflow reference is docs/01-agent-workflow.md.

Architecture

Company and board CSVs are the configuration source of truth. The crawler keeps operational state in its own Postgres database, uses Redis for scheduling, publishes searchable documents to Typesense, and stores full descriptions in S3-compatible object storage. The web app owns authentication, watchlists, and application-tracker data in a separate Postgres boundary.

%%{init: {"flowchart": {"rankSpacing": 24}}}%%
flowchart TD
    Requests["Company requests"] --> Agent["Coding-agent PR"]
    Agent --> CSV["companies.csv + boards.csv"]
    CSV --> Runtime["Crawler runtime<br/>sync · Redis · HTTP/browser workers · crawler Postgres"]
    Runtime --> ReadLayer["Published read layer<br/>Typesense · S3-compatible descriptions"]
    ReadLayer --> Web["Next.js<br/>web app · REST API · MCP"]
    Web <--> WebDB["Web Postgres"]

Start with the maintained documentation index, then read the system overview, crawler architecture, and Typesense reference.

Run it yourself

Job Seek is self-hostable, but the repository currently expects operator-managed services. The root docker-compose.yml is development scaffolding for Postgres and Typesense, not a complete one-command production deployment.

Prerequisites

• Python 3.13+ and uv
• Node.js 22+ and pnpm 10
• A crawler-owned Postgres database
• A separate web-owned Postgres database; both databases may run on the same server
• Redis
• Typesense
• S3-compatible object storage for job descriptions

The important environment boundaries are:

After provisioning those services, create apps/crawler/.env.local (using its .env.example as a starting point) and apps/web/.env.local, then initialize in this order:

git clone https://github.com/colophon-group/jobseek
cd jobseek
corepack enable
pnpm install

# Web-owned database
cd apps/web
pnpm db:migrate

# Crawler-owned database and derived stores
cd ../crawler
uv sync
uv run playwright install chromium-headless-shell  # required by run-browser
uv run alembic -c src/migrations/alembic.ini upgrade head
uv run crawler setup-typesense
uv run crawler sync

A complete crawler deployment keeps four process roles running:

uv run crawler run          # HTTP workers
uv run crawler run-browser  # Playwright workers
uv run crawler export       # Postgres → Typesense CDC
uv run crawler drain        # descriptions → object storage

Start the frontend separately from apps/web:

pnpm dev                    # http://localhost:3000

Production operators should use scoped Typesense keys and read the deployment and recovery runbooks linked from docs/README.md.

What's in the repo