Define what you actually want, vet real jobs before chasing them, write honest applications from your own evidence, and track every outcome — all on your own machine, with your own AI.
It onboards you once — your preferences, your evidence, your comp floor — and then works the whole loop: finding jobs, vetting them, writing honest applications, drafting recruiter messages, prepping you for interviews, and learning from every outcome. The agent does the work. You make the calls.
“It reads the actual job before you chase it.”
Every posting gets a verdict: keep it or cut it, how well it fits, whether the money works, and what to do next. Judged against your constraints, not a keyword match.
Each step builds on the last. Nothing is skipped, nothing is faked.
Answer a handful of questions. The roles you want, the least you'll take, where you'll work, what you won't put up with. Your answers live in plain files you can read. No black box.
Searches the boards you pick. Reads the whole posting, not the headline. Weighs it against your comp floor, your location, your dealbreakers — then tells you to go for it, skip it, or take a closer look. No spray-and-pray.
Tailors your résumé and cover letter only from your real evidence bank. Refuses to invent facts. Every claim is backed by something you actually did.
Drafts recruiter messages and follow-ups. Builds story banks for interviews. Coaches comp negotiation. Tracks every outcome and gets smarter each run.
Honest tailoring isn't a limitation — it's a design choice. Applications that overclaim don't hold up in interviews. Artifacts built from your real evidence do.
You build a bank of real things you did — projects, metrics, decisions. Every tailored artifact draws only from that bank.
If a claim isn't in your evidence, it won't write it. No hallucinated roles. No inflated titles. No fake metrics.
It reorders, reframes, and emphasises what genuinely fits the role. That's tailoring. That's honest.
Tailoring only runs after the gate clears. You never write a cover letter for a job that doesn't fit your constraints.
Run Rolester on your own machine and your résumé, comp numbers, evidence bank, and full pipeline stay on your laptop. No account to create, no telemetry, nothing syncing in the background.
Job searches are personal. Your comp floor, your reasons for leaving, your backup options — run Rolester yourself and all of it stays on your machine, full stop.
No signup, no telemetry, nothing phoned home. The self-host version runs on your machine and uses the AI CLI you already have. What you put in stays in.
Rolester is an agent runtime. It doesn't lock you into a model or a subscription. Bring whatever AI CLI you already use — Claude Code, Codex, or anything else on your PATH.
Rolester scaffolds the workspace, installs skills, starts the dashboard, and hands off to a supported agent CLI. Model choice, cost, and local data stay under local control.
It all ships together, and none of it is built for one industry. A nurse, an engineer, and a driver each answer onboarding their own way and get the same loop.
Free to self-host, open source (MIT). No wizard, no signup — two commands and the agent takes it from there.
npm install -g @anthropic-ai/claude-code claude.com/claude-codenpm install -g @openai/codex github.com/openai/codexPaste a job posting and say “evaluate this” to kick off the loop. Or try the bundled sample under examples/sample-jobs/.
Free. Local. Honest. One command to get going. Run it yourself and your data stays put.
# one command, then the agent takes it from here