v0.17.0 · Apache-2.0 · Local-first

The career layer that lives inside your coding agent.

Valor turns Claude Code, Codex CLI, or Cursor into a career coach. It plans your day, coaches your reviews, captures what you did, and at review time writes the draft you'd otherwise reconstruct from memory. No new tool to open.

Try the demo Star on GitHub
  • 0runtime deps
  • 3agents supported
  • 100%local storage

What it does

Three moments in your week.

Eleven on-demand workflows plus ambient coaching, all triggered by plain language. Everything runs against a local evidence store — no dashboards, no telemetry.

Every morning

Tickets, your review queue, today's calendar and yesterday's carry-forward, ranked against the goals in your own 1:1 doc — then fitted to the day's real gaps, deep work in the deep block, prep reserved before the meetings you present at.

start my day wrap up

While you work

Senior-level review before you hit "request changes", a design doc with the trade-offs written down, and a PR turned into a diagram a cold reviewer can follow. Coaching arrives unprompted, tied to your own career framework.

help me review #418 design doc for PROJ-231

At review time

A review cycle's captured evidence becomes a self-reflection draft mapped to your competencies — phrased from the recorded status of each item, with work you forgot to log surfaced and a list to confirm before you submit instead of overclaiming.

prep for 1:1 half-year reflection

See all eleven with real output, in a session you can talk to

How it works

Install → setup → use. That's it.

  1. 01

    Install once

    Drops the skills, agents, and rules into each AI agent's config directory.

    curl -fsSL …install.sh | bash -s -- --clone

  2. 02

    Configure in your agent

    Say "set up valor". Paste your career ladder — or start from a bundled example. Pick your level. Done.

    set up valor

    paste a ladder, start from an example, or generate one from your title

  3. 03

    Use it as you work

    Talk to your agent like normal. Ambient coaching runs alongside.

    start my day wrap up prep for 1:1

Local-first

Your career data stays on your machine.

Apache-2.0, plain Python and Markdown, no backend and no telemetry. Valor reaches only the tools your agent already has — PRIVACY.md draws the exact trust boundary.

Hosts

Claude Code Codex CLI Cursor

Runs on

macOS Linux Python 3 stdlib only

Your data

~/.valor SQLite + plain text no telemetry no backend

Install

One command, then you can forget it happened.

The installer drops Valor's skills, agents, and rules into ~/.claude/, ~/.codex/, and ~/.cursor/. After that, your agent is the interface.

curl -fsSL https://raw.githubusercontent.com/yihanzhu/valor/main/install.sh | bash -s -- --clone
  • Read first The bootstrap is fetched from main. It's a short shell script — read it before piping to bash: curl -fsSL …/main/install.sh | less.
  • Per host Only need one agent? Pass --target claude-code, codex, or cursor.
  • Updates Checks for new releases (tagged vX.Y.Z, never main) and notifies you — updating is manual. Pin with git -C ~/.valor/repo checkout vX.Y.Z.
  • Local state Career framework, evidence, and carry-forward notes live in ~/.valor/. Valor never sends them anywhere; what your host assistant does with your prompts is governed by it, not by Valor — PRIVACY.md is explicit about that line.