mycelium

agent
Guvenlik Denetimi
Gecti
Health Gecti
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Community trust — 10 GitHub stars
Code Gecti
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Gecti
  • Permissions — No dangerous permissions requested

Bu listing icin henuz AI raporu yok.

SUMMARY

Organize your AI sessions, and carry what they know into the next one.

README.md

Mycelium

npm version
npm downloads
CI
coverage
license: MIT
node
Ask DeepWiki

Organize your AI sessions, and carry what they know into the next one.

You create a lot of AI sessions while working, and most of them get lost the moment they end. Mycelium organizes the ones worth remembering and extracts their knowledge, insights, and context, so the next one already knows what mattered.

  • Capture — Claude Code, Codex, Kiro, and OpenCode sessions land in one place, automatically.
  • Organize — grouped by the project, case, or service they belong to.
  • Learn — what repeats across a folder's sessions becomes durable knowledge.
  • Reuse — your next session inherits that knowledge, insights, and context automatically when started from Mycelium.

Local-first, model-agnostic, human-controlled.

Demo

mycelium demo — a Cloud Support Engineer's cases, organized and learned from

mycelium demo walks through the same loop interactively — try it with
npm install -g @kevinprk/mycelium && mycelium demo, or see it as a
Software Engineer / Solutions Architect instead of a Cloud Support
Engineer by picking a different persona when it asks.

Requirements

Mycelium uses your own LLM usage. organize, autotag, knowledge,
and split suggestions call your own claude/codex CLI — including
automatically in the background while mycelium is open. Disable with
MYCELIUM_NO_AUTOSTART=1, or tune it in docs/cli.md.

  • Node.js ≥ 22.16, git
  • AI agents: claude / codex / kiro-cli / opencode

Install

npm install -g @kevinprk/mycelium   # via npm
# or
brew install krapie/tap/mycelium    # via Homebrew

To hack on it, clone instead:

git clone https://github.com/krapie/mycelium.git
cd mycelium && npm install && npm link

Getting Started

mycelium        # launch the TUI — first run offers a guided tour
mycelium demo   # try it risk-free first — separate store, your real data untouched

Sessions from Claude Code, Codex, Kiro, and OpenCode are captured automatically in the
background — no manual scan needed. Day to day it's a simple loop, the
Context Flywheel: s capture → o organize → w learn → n start the
next session with everything it needs. Most of that already runs by itself
in the background; pressing these keys is mostly reviewing and confirming
what's already waiting for you.

mycelium demo walks through that same loop against mock sessions, in your
choice of language and persona, then hands off straight into your real
sessions once you finish it — an early Esc bail just exits, your actual
~/.mycelium store untouched either way.

Display language defaults to English — switch any time with
mycelium lang <en|ko>, the in-TUI l key, or the language picker shown on
first launch.

Skipped npm link? Use node src/cli.js <command> instead.

Mycelium stores everything on your machine (~/.mycelium/) with no
server or telemetry — safe to use inside restricted orgs that already
permit claude/codex. Just note that its LLM calls
(organize/autotag/knowledge) do carry session content to the provider,
the same as any direct CLI use.

Learn More

Full guide in docs/:

  • How It Works: the loop, and why it holds together, in two diagrams
  • TUI — the 3-column interface + every keyboard shortcut
  • Learn/Reuse loop — how sessions pass knowledge forward
  • Handoff — continuing work across agent CLIs
  • CLI reference — every subcommand, for scripting
  • Architecture — data location, design principles, status
  • Feature catalog — every capability, with test-coverage status

Cleanup (experimental stage)

mycelium cleanup            # safe: removes Mycelium's own LLM-call sessions + empty folders, rebuilds the index
mycelium cleanup reset --yes # irreversible: wipes ~/.mycelium entirely

tidy/folders/archive/index/reset never touch your original agent
logs, so mycelium scan can always rebuild — see
docs/cli.md for the full breakdown.

Contributing

See CONTRIBUTING.md for dev setup, lint/test, and how
to add a new AI agent CLI.

Working in this Repo with an AI Agent

Using Claude Code, Codex, Cursor, or similar on Mycelium itself? Point it at
AGENTS.md first — a dense reference covering conventions, key
bindings, and the contributing workflow. (Claude Code doesn't read AGENTS.md
on its own, so CLAUDE.md in this repo's root just imports it.)

Yorumlar (0)

Sonuc bulunamadi