SSyncDev

Worklife MCP Server

A local job-search workspace for agents: vacancy search, CV matching, application pipeline, interviews and offers.

Local server

What is the Worklife MCP Server?

Job hunting produces a lot of state. Saved vacancies, three versions of a CV, recruiter emails, interview dates, two offers with different payment terms. The Worklife MCP server gives an agent a structured place to keep all of it, plus tools to go and find more, while your personal data stays in a workspace on your own machine.

Who it suits

It is aimed at an individual job seeker who already works inside an MCP client such as Claude Code, Cursor, Codex or OpenCode. The model runs in your client; Worklife supplies the tools and storage. An optional built-in runner can drive its own loop if you point it at an OpenAI-compatible endpoint, but nothing is configured or called by default.

The working toolkit

Tools in the Worklife MCP server are grouped by stage of the search:

  • Market — market_collect, market_snapshot and market_report gather explicit queries and report salary medians and quantiles, kept separate by currency, pay period and gross or net
  • Matching — recommendations_report, cv_match and recommendation_save compare vacancies with your stated conditions and experience facts
  • Pipeline — pipeline_update, inbox_sync and conversation_update track stages, next steps and linked replies
  • Interviews and offers — interview_brief, interview_save, offer_save and offer_compare
  • Hygiene — employer_block keeps a personal blocklist; automation_configure schedules recurring checks

Recording an offer does not accept it, and saving an interview sends no invitation. CVs import from TXT, Markdown, PDF with a text layer, or DOCX, and export to Markdown, HTML or PDF.

Getting it running

Python 3.12 or newer and uv are needed. The quickest route is the setup command with a target for your client, for example --target claude-code or --target cursor. It builds a persistent runtime, installs Chromium for browser automation and prepares a career folder in your home directory. Open that folder in your client and begin with recruiter_start.

For an existing stdio client you can instead run the pinned worklife-recruiter mcp entry through uvx, which stores data in ~/.worklife. Browser features then need the browser extra and Chromium installed separately.

Before you rely on it

Accounts on job sites, email (SMTP and IMAP) and a Telegram bot are configured by you; there is no universal access to platforms. Passwords created by credential_create live in the system keychain, and there is no unencrypted fallback. A shared, signed knowledge base of employer observations syncs in the background; only employer information goes there, not your profile. Much of the detailed documentation is in Russian. Related tools include Adzuna Job Search and the wider HR and recruiting category.

Recorded tools

Check the linked publisher documentation and your connected client's tool list. Names and availability can vary with the server version and account permissions.

ToolWhat it does
recruiter_startStarting point for a session; the agent asks about your job-search goals.
market_collectCollect vacancies for explicitly chosen search queries.
market_snapshotSave an immutable sample of collected vacancies.
market_reportReport salary coverage, medians, quantiles and source limits, split by currency, period and tax basis.
job_saveSave a vacancy found through the browser so it can join a sample.
recommendations_reportCheck vacancies against your conditions and experience facts.
cv_matchCompare CV versions against a vacancy.
pipeline_updateUpdate an application's stage, next step and deadline.
inbox_syncStore incoming email and link replies to applications.
interview_briefPrepare questions and materials for an interview from the vacancy and your experience.
offer_compareCompare saved offers, including terms and payment timing.
cv_exportExport a CV to Markdown, HTML or PDF.
employer_blockAdd a company to your personal blocklist so it is excluded from results and contact.
automation_configureSchedule recurring search refreshes, inbox checks and reviews.

How to install the Worklife MCP Server

{
  "mcpServers": {
    "worklife_recruiter": {
      "command": "uvx",
      "args": [
        "--python",
        "3.12",
        "--from",
        "worklife-agent==0.5.2",
        "worklife-recruiter",
        "mcp"
      ]
    }
  }
}

Needs uv and Python 3.12+. Data is stored in ~/.worklife. Browser automation additionally needs the browser extra and Chromium.

Configuration

VariableDescriptionRequired
WORKLIFE_HOMEOverrides where personal data is stored (default ~/.worklife).Optional
WORKLIFE_NETWORK_OFFLINESet to 1 to stop update checks against PyPI.Optional
WORKLIFE_AGENT_BASE_URLEndpoint of an OpenAI-compatible service for the optional built-in agent runner.Optional
WORKLIFE_AGENT_MODELModel name for the optional built-in runner.Optional
WORKLIFE_AGENT_API_KEYAPI key for the built-in runner's endpoint, if it needs one.Optional
WORKLIFE_NETWORK_GITHUB_TOKENToken used to publish employer observations to the shared knowledge base; an existing gh auth login also works.Optional

Frequently asked questions

In a local workspace, by default ~/.worklife. The shared knowledge base only carries employer information, not your profile, CVs or correspondence.