tjqscott

Dory — AI skill for Claude Code

AI community

Dory treats hosted LLM interfaces as disposable tools by offloading state to your local environment and enforcing atomic sessions.

How to install Dory

This entry records only its repository, not the path inside it, so there is no exact command to give. Open tjqscott/dory and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Dory does

Dory treats hosted LLM interfaces as disposable tools by offloading state to your local environment and enforcing atomic sessions.

Alternatives in AI

  • Hindsight — State-of-the-art long-term memory for AI agents by Vectorize 6.7k ★
  • Update Branch Name — by giselles-ai - Updates branch names with proper prefixes and formats, enforcing naming conventions, supporti 502 ★
  • Gensee Crate — Run AI harnesses like Claude Code, Codex, Copilot, Antigravity, and Omnigent in disposable workspace forks 123 ★

README

Dory

Dory

Stateless context management for LLM sessions.

Stars Release 3.0.0 Works with Claude, ChatGPT, Codex, Local LLMs MIT license


Every new chat starts blind. Long threads drain quotas and dilute context, and closing the tab loses the project. Dory keeps the memory outside the chat, so every session can start fresh and still know what was decided, what was tried, and why.

It comes in two editions, one for each way you work with a model.

🌐 Browser edition 💻 Code edition
For any hosted chat: Claude, ChatGPT, Gemini, local models Claude Code (plugin) and Codex (AGENTS.md)
Memory posts the model writes, kept in one file you own a wiki/ of small linked pages in your repo
Install download one HTML file /plugin install dory@dory
Enforced by the page, which rejects malformed posts hooks, which lint the wiki before a session can finish

🌐 Browser edition

  1. Open `browser/single-file/dory.html` in your browser.
  2. Click Copy context and paste it into a fresh chat. It holds the directives, the posting format, the pages you tick and, if you read it, your project folder's tree and data previews. The size is measured, not guessed.
  3. Work. At the end, the model emits posts.
  4. Paste its whole reply back, click Add posts, then Save.

Posts are append-only.