🪨 why use many token when few token do trick — Claude Code skill that cuts 65% of tokens by talking like caveman
Task tracker for Claude Code and other agents
Dex tracks agent work as structured tasks with a description, context, and result. It keeps that state in JSONL so agents can pick up where they left off and teams can merge changes cleanly.
Builders who run agent-led work across multiple sessions and want the same task context to persist.
You can keep complex work organized, resumed, and recorded without retyping the background each time.
What it does
Structured task records
Each task stores a short description, background context, and final result.
Persistent session memory
Agents can revisit prior work and continue from the recorded task state.
Git-friendly storage
Tasks are saved as JSONL, which makes them easy to version and merge.
Agent commands and plugin support
The repo ships with agent-facing command files and a local plugin setup for Claude Code.
How to get it
- 1Install dex globally
npm install -g @zeeg/dex
- 2Or use the skills installer for your AI agent (Claude Code, OpenCode, Codex, Cursor)
npx skills add dcramer/dex
- 3Use natural language or slash commands
> Use Dex. Work on the next logical task. When done, run code-simplifier and commit.
- 4Run
/dex create a task to refactor the authentication module
- 5Run
git clone git@github.com:dcramer/dex.git cd dex pnpm install && pnpm build pnpm link # Makes 'dex' command available globally
README
dex
Task tracking for AI agents. Persistent memory for complex, multi-session work.
Why dex?
Structured tasks, not simple todos. Each task captures full context:
- Description: One-line summary (like an issue title)
- Context: Background, requirements, approach (like an issue body)
- Result: Implementation summary, decisions, outcomes (like a PR description)
Built for agent coordination. Agents break down work, track progress, and record results that persist across sessions.
Git-friendly storage. JSONL format (one task per line) enables collaboration, versioning, and conflict-free merges.
Quick Start
Install dex globally:
npm install -g @zeeg/dex
Or use the skills installer for your AI agent (Claude Code, OpenCode, Codex, Cursor):
npx skills add dcramer/dex
Use natural language or slash commands:
> Use Dex. Work on the next logical task. When done, run code-simplifier and commit.
/dex create a task to refactor the authentication module
Contributing
Development Setup
git clone git@github.com:dcramer/dex.git
cd dex
pnpm install && pnpm build
pnpm link # Makes 'dex' command available globally
Development Cycle
pnpm dev # Watch mode - auto-rebuild on changes
Testing Locally
Test the plugin locally without installing from the marketplace:
claude --plugin-dir plugins/dex
This loads the plugin from the local directory. Restart your AI agent to pick up changes.
Manual Build
# Make changes...
pnpm build
pnpm test
Releasing
Releases are automated via GitHub Actions when a version tag is pushed.
# Run tests, build, and bump version (creates git tag)
pnpm release patch # 0.1.0 → 0.1.1
pnpm release minor # 0.1.0 → 0.2.0
pnpm release major # 0.1.0 → 1.0.0
# Push commit and tag to trigger publish
git push --follow-tags
The workflow runs tests and publishes to npm with provenance.
Files in the repo
- .claude
- .claude-plugin
- .dex
- .github
- docs
- plugins
- specs
- src
- .gitignore
- AGENTS.md
- CLAUDE.md
- dex.toml.example
- LICENSE
- package.json
- pnpm-lock.yaml
- README.md
- TESTING.md
- tsconfig.json
- vitest.config.ts
- vitest.setup.ts
Discussion (0)
Ask about usage, or say what you built with itSign in to join the discussion.
No comments yet. Be the first to say what this is good for.
More tools
The best-benchmarked open-source AI memory system. And it's free.
Orca is the ADE for working with a fleet of parallel agents. Run any coding agent with your own subscription. Available on desktop, mobile and remote runtime.

A cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode, OpenClaw, Grok Build & Hermes Agent. Only official website: ccswitch.io
Never stop coding. Free MIT AI gateway: one endpoint, 352 providers (150+ free), 1200+ models Kimi, Claude, GPT, Gemini, GLM, DeepSeek, MiniMax. Works with Claude Code, Codex, Cursor, OpenCode, Cline & Copilot. Quota-aware auto-fallback, RTK+Caveman compression saves 15-95% tokens, MCP/A2A, Desktop/PWA. Built by 550+ contributors
Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.