tinker-cookbook
Post-training with Tinker
Tinker Cookbook is a Claude Code plugin that teaches Claude how to use the Tinker API for post-training and fine-tuning language models. It provides three specialized skills—research, debug, and inkling—that guide Claude through supervised fine-tuning, reinforcement learning, DPO, distillation, and evaluation tasks. Developers use these skills to accelerate setup and troubleshooting of LLM customization workflows, with skills automatically triggering based on conversation context or invoked via slash commands.
Key Features
Use Cases
- 01Fine-tuning language models on conversational datasets with supervised learning guidance
- 02Implementing reinforcement learning for mathematical reasoning and competitive programming tasks
- 03Setting up multi-stage RLHF pipelines including reward model training and policy optimization
- 04Debugging slow training runs, renderer mismatches, and API integration errors with Claude's assistance
- 05Configuring knowledge distillation from single or multiple teacher models
- 06Evaluating trained models against standard benchmarks with verified scoring configurations
Related Plugins
View moreandrej-karpathy-skills
A single CLAUDE.md file to improve Claude Code behavior, derived from Andrej Karpathy's observations on LLM coding pitfalls.
claude-mem
Persistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode + More
Understand-Anything
Graphs that teach > graphs that impress. Turn any code into an interactive knowledge graph you can explore, search, and ask questions about. Works with Claude Code, Codex, Cursor, Copilot, Gemini CLI, and more.
rtk
CLI proxy that reduces LLM token consumption by 60-90% on common dev commands. Single Rust binary, zero dependencies
tinker-cookbook — FAQ
What is the Tinker Cookbook Claude Code plugin?+
It's a set of three specialized skills that teach Claude Code how to work with the Tinker API for fine-tuning and post-training language models. The skills cover experiment planning, debugging, and working with Inkling models.
How do I install the Tinker Cookbook plugin in Claude Code?+
Run '/plugin marketplace add thinking-machines-lab/tinker-cookbook' in Claude Code, then open the Discover tab and install the 'tinker' plugin from the list. The skills become immediately available for use.
Do I need a Tinker API key to use this plugin?+
Yes, you must sign up at auth.thinkingmachines.ai, create an API key from the Tinker console, and export it as the TINKER_API_KEY environment variable. The plugin guides Claude in using this API for training operations.
Which AI clients work with this plugin?+
This plugin is designed specifically for Claude Code and uses Claude Code's native skill system with slash commands and automatic context-based triggering.
Is the Tinker Cookbook plugin free to use?+
The plugin itself is open-source and free, but using the underlying Tinker API requires a paid account. Pricing varies by model and is detailed in the Tinker documentation.
Can Claude automatically use Tinker skills without slash commands?+
Yes, the skills trigger automatically based on conversation context. For example, asking Claude to 'set up SFT training' will load the appropriate skill without needing to type a slash command.
How do I install tinker-cookbook?+
Open the source repository on GitHub and follow its README. tinker-cookbook is a plugin — MCP Agents Market links you directly to the official repo.
Is tinker-cookbook free?+
tinker-cookbook is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.