Work that you love, so you never work a day in your life.
LoveWork is a work-discovery agent, personal guide to achieving your life's work. Movie stars and successful creatives have human agents who look continuously for interesting projects, people, and opportunities. LoveWork makes that kind of representation available to everyone.
It is not merely a job-search engine. A job is work done for someone else for money; worthwhile work can also be a collaboration, commission, research project, company, challenge, residency, or direction that does not yet have a conventional title.
LoveWork builds a living model of a person's experience, circumstances, constraints, interests, and possible futures. It searches widely, assesses opportunities in that context, remembers decisions and outcomes, and improves its future recommendations.
This repository is the early public face of an active work in progress. The ideas and design are public before the product is finished because they are already useful to discuss, test, and build upon.
- Manifesto — why LoveWork should exist.
- Why LoveWork — the problem with episodic job search.
- Profile model — representing a person without reducing them to a CV.
- Intelligence layer — the longitudinal judgment loop at the centre of the system.
- Architecture — the current technical shape.
- Roadmap — what exists and what comes next.
- Privacy — the boundary between open software and private human data.
- Synthetic example — what a future report could look like, using fictional data.
public sources -> discovery and crawling -> opportunity registry
|
private profile -> contextual assessment --+
|
decisions and outcomes -> learning loop -> reports and suggestions
The working prototype already crawls public career and opportunity sources, tracks their lifecycle, scores them against layered candidate profiles, checks prior contact, and writes a browsable local wiki. The public release will grow from this documentation-first foundation toward a reusable engine.
The working Python engine is included in lovework-agent/,
along with source adapters, the dashboard/MCP surface, and its test suite. The
bundled profiles/example/ is fictional and exists only
to show the profile contract; active profiles, application records, crawl
results, and decision ledgers remain private.
uv venv venv
uv pip install --python venv/bin/python -e "lovework-agent[dev]"
cd lovework-agent
../venv/bin/python -m pytest tests/ -qRunning a live crawl additionally requires your own LLM and optional web/Gmail
credentials through environment variables. See
lovework-agent/README.md for the engine shape.
Early and experimental. Interfaces, terminology, and implementation will change. The present repository exists so LoveWork has a public address while the project is still being shaped.
Software is licensed under Apache License 2.0. The manifesto and documentation are licensed under Creative Commons Attribution 4.0 International. See LICENSE.