# Enzyme

> Enzyme makes an existing markdown workspace operable by coding agents. It compiles folders, tags, wikilinks, dates, frontmatter, and prose into source-grounded retrieval so an agent can prep from the user's own notes, project logs, meeting captures, decisions, and observations.

Primary user: a markdown/Obsidian power user using coding agents. Enzyme is not a hosted chat-memory database by default and does not require a proprietary hidden graph. The user's markdown remains the source of truth.

## Start here

- [Full Agent Guide](https://memory.enzyme.garden/llms-full.txt): Current agent-readable setup, trust, runtime, and proof guide
- [Workspace Setup Skill](https://raw.githubusercontent.com/useenzyme/enzyme/main/skills/enzyme-workspace-setup/SKILL.md): First-time setup contract agents should read before scan/init
- [Enzyme in 90 seconds](https://memory.enzyme.garden/docs/enzyme-in-90-seconds/): What Enzyme is and who it is for
- [First successful run](https://memory.enzyme.garden/docs/first-successful-run/): Project/meeting-prep proof with source receipts
- [Quickstart](https://memory.enzyme.garden/setup): Install the CLI and runtime instructions
- [What Enzyme is / is not](https://memory.enzyme.garden/docs/what-enzyme-is-is-not/): Category boundaries and non-goals
- [Data flow & trust](https://memory.enzyme.garden/docs/data-flow-and-trust/): What gets written, what may call a provider, and how to undo it
- [Supported runtimes](https://memory.enzyme.garden/docs/supported-runtimes/): Codex, Claude, Pi/generic agents, Hermes/OpenClaw status
- [Troubleshooting](https://memory.enzyme.garden/docs/troubleshooting/): Weak first results, provider errors, stale indexes, runtime setup

## Concepts

- [How it works](https://memory.enzyme.garden/docs/how-it-works/): scan → init → petri → catalyze → refresh
- [Memory is not search](https://memory.enzyme.garden/docs/memory-is-not-search/): When to use Enzyme vs exact search or RAG
- [Catalysts](https://memory.enzyme.garden/docs/what-catalysts-are/): Source-grounded questions generated from the workspace
- [What the agent sees](https://memory.enzyme.garden/docs/what-the-agent-sees/): How agents use petri/catalyze output
- [Capture memory as artifacts](https://memory.enzyme.garden/docs/how-agent-memory-should-be-captured/): Conservative source-linked writeback policy
- [Benchmarks](https://memory.enzyme.garden/docs/benchmarks/): What 8ms query-time retrieval measures
- [Comparisons](https://memory.enzyme.garden/docs/comparisons/): Enzyme vs search, RAG, managed memory APIs, coding-agent memory tools

## Status notes

- Local CLI path is the primary docs path.
- Catalyst generation during init/refresh may send selected excerpts to the configured/default provider.
- Query-time retrieval from the compiled index is local and does not call an LLM.
- Speed claims such as 8ms query-time retrieval are claimable when scoped to the compiled local retrieval path.
- Hermes and OpenClaw support are experimental.
- Product-corpus SDK is private beta.