Codex / Claude / Pi

Use this path for Codex, Claude Code, Pi, and other agents that read AGENTS.md or local skills. The Enzyme skill tells your agent how to set up and use Enzyme from inside your markdown workspace.

If you are comparing runtimes, start with Supported runtimes. The short version: Codex and Claude Code have dedicated install targets, Pi uses the generic .agents path, and Hermes/OpenClaw are experimental.

Quick start

Install the CLI:

curl -fsSL byenzyme.com/install.sh | bash

Then run the install command for your agent from the vault root:

cd /path/to/your/vault

enzyme install codex      # Codex, Pi, and generic .agents readers
enzyme install claude     # Claude Code

Then ask your agent:

Set up Enzyme for the Markdown notes folder I have open. If you cannot identify it, ask me for its path.

Install Enzyme if needed using https://raw.githubusercontent.com/byenzyme/enzyme/main/install.sh, and install the instructions for this agent. Read and follow the installed enzyme-workspace-setup skill.

You may use Enzyme's online service to suggest settings from my notes and prepare search. Show me what you propose to include or exclude, and let me correct it before building the local search index.

Then answer one useful question using my notes, with links to the sources, and suggest a question I can ask next. Keep my existing notes in place and unchanged. Ask separately before any cleanup, model download, or recurring task.

The install command writes durable agent instructions plus two skills for Codex/Claude-style runtimes: enzyme-workspace-setup for first setup and repair, and enzyme for normal retrieval, refresh, and note-writing. The agent should read the setup skill before running petri, compile, or init.

Provider safety is part of that setup: default enzyme init and enzyme refresh use Enzyme hosted credits/auth and ignore inherited LLM env keys. Agents should not silently spend personal OpenAI/OpenRouter keys. No shell-variable cleanup is needed; use --use-env-llm only when you intentionally want your own provider.

It should preserve existing Obsidian or markdown conventions: folders, tags, wikilinks, daily notes, people pages, project notes, and frontmatter. It should not impose a hidden memory architecture.

What the skill should prove

A good first run ends with source-grounded project context, not just a green command.

Ask for something concrete:

Prep me for the next meeting on <project>. Cite the notes that show unresolved commitments, decisions, or source context.

The agent should use:

enzyme petri
enzyme catalyze "prep for next meeting on <project> unresolved commitments decisions"

A useful result names source files, quotes excerpts, and explains why those notes matter. If the result is generic, ask the agent to inspect config, exclude noisy folders, add clearer markdown handles, or run enzyme refresh after repairs.

Terminal-only fallback

If you prefer to run setup directly in a terminal:

enzyme compile -v
# Review the .enzyme settings file at the path printed by compile.
enzyme init
enzyme petri
enzyme catalyze "prep for the next meeting on <project>"

When to use it again

Use the skill when you want the agent to:

  • explain what Enzyme sees in the vault;
  • improve entity/folder selection without adding noise;
  • check whether people, projects, clients, meeting notes, or Readwise folders are being read correctly;
  • decide whether an external folder should be added with enzyme apply;
  • write compact, source-linked markdown observations that future Enzyme refreshes can retrieve.

For normal sessions, the agent can use:

enzyme refresh --quiet
enzyme petri --query "user question"
enzyme catalyze "query using petri vocabulary"

Writeback policy

Do not ask the agent to “remember everything.” A good Enzyme writeback is an ordinary markdown artifact that captures durable context:

  • decisions;
  • open loops;
  • commitments;
  • project state changes;
  • durable preferences with evidence;
  • useful people/company context with source links.

The agent should not write raw tool output, secrets, one-off commands, generic session summaries, or edits to canonical docs/roadmaps without review. See Capture memory as artifacts for the policy.

Claude Cowork

In Claude Cowork, install the Enzyme marketplace/plugin path from the setup page and use /enzyme for the guided review flow. The CLI installer does not install that app plugin; it installs the local enzyme binary and agent instructions for local agents.