Installation

Installation

One command

brew install coder8124/tap/logos-mcp && logos setup    # macOS and Linux
npm i -g @noeton/logos && logos setup                  # Windows

That is the whole install. You need no Go toolchain, no clone and no build.

• The Homebrew formula is logos-mcp from the coder8124/tap tap. A bare brew install logos installs Logos Bible Software, a different product. The command it installs is still logos.
• Update with brew upgrade logos-mcp. Hosts are wired to brew's opt path, which survives the upgrade.
• Windows has no Homebrew, so there you install @noeton/logos from npm. It carries a prebuilt binary for your platform, about 5 MB.
• Prefer an installed binary over npx where you can. npm asks the registry for the package on every npx start. Without a connection it waits and then fails.

Other routes

# From source, if you have Go
git clone https://github.com/Coder8124/logos ~/src/logos && cd ~/src/logos
go build -o bin/logos ./cmd/logos && ./bin/logos setup

# A release binary: github.com/Coder8124/logos/releases
# On macOS, clear the browser's quarantine flag once before the first run:
xattr -d com.apple.quarantine ./logos

What logos setup does

$ logos setup

  vault      /Users/you/logos   (created)
  runtime    Ollama at http://localhost:11434/v1
  embedding  nomic-embed-text ✓
  model      gemma3:4b ✓

  index      0 notes, 0 edges

  hosts
    each of these will be pointed at:
      /opt/homebrew/opt/logos-mcp/bin/logos mcp serve
      LOGOS_VAULT=/Users/you/logos

    Claude Code      →  claude mcp add --scope user
    Cursor           →  ~/.cursor/mcp.json
    Codex            →  codex mcp add

  wire 3 host(s)? [Y/n] y
  ...
  checking it works
    handshake        ✓  server answered initialize
    tools            ✓  continuity tools advertised
    round trip       ✓  checkpoint written and recovered through resume
    vault            ✓  written to /Users/you/logos, and cleaned up

1. Picks a vault. It uses ~/logos unless you pass --vault DIR.
2. Finds a local model runtime (Ollama, LM Studio, Jan or Msty) and offers to pull anything missing. A runtime is optional; see below.
3. Runs the first index.
4. Shows which agents it would wire and asks before touching any of them.
5. Proves it works. Setup acts as the host for one round trip: it writes a checkpoint, reads it back through resume, confirms the markdown landed in the vault it configured, then deletes it. A host pointed at the wrong vault passes a handshake while knowing nothing, and that is the failure this step catches.

Useful flags:

Flag	Effect
--dry-run	Print the whole plan and write nothing
--host cursor	Wire exactly one host
--vault ~/notes	Use a different vault folder
--no-hosts	Make the vault and index, but wire no agents
--yes	Don't ask

Setup is safe to run twice. It updates an existing entry rather than adding a second one, it leaves the other MCP servers in a config file alone, and it saves a .logos-backup beside anything it edits. logos mcp install re-runs just the wiring, and logos doctor --integration re-runs the round-trip check.

Claude Code

When Claude Code's CLI is on your PATH, setup installs the Logos plugin and updates it whenever it is older than your logos binary. The plugin adds:

• the MCP server
• a SessionStart hook that puts the last handoff in front of the model before it does anything
• a pre-edit hook that shows a recorded dead end beside an edit to a file it names
• the /handoff command
• the context-connect and continuity skills

To install the plugin by hand instead, from inside Claude Code:

/plugin marketplace add Coder8124/logos
/plugin install logos@logos

The plugin runs the logos binary but does not include one, so install logos first.

Do I need a model?

No. Every continuity tool (checkpoint, resume, note_progress, before_you_try, why) is markdown and SQL, and works with nothing installed. Search falls back to BM25 keyword ranking. For code, which is mostly identifiers, error strings and paths, that is the right tool.

What a local runtime adds:

Piece	Size	What it gives you
Embedding model (nomic-embed-text)	~274 MB	Search that matches by meaning. before_you_try catches the same idea worded differently
Chat model (LOGOS_MODEL, default qwen3.6)	4–24B	logos search --answer and the nightly logos memory dream only. No MCP tool uses it

Set LOGOS_RUNTIME=off to run with no model at all, or set it to an OpenAI-compatible base URL to skip auto-discovery.

Checking it

logos doctor                  # vault, index, hosts, privacy, review queue
logos doctor --integration    # prove each host can actually reach this vault
logos doctor --verbose        # adds runtimes and model tiers

Uninstalling

logos setup --uninstall       # or: logos mcp uninstall [--host NAME]

This removes the logos entry from each host's config, backs up anything it changed, and says what it removed. It never touches the vault. Then remove the binary the way you installed it (brew uninstall logos-mcp, npm uninstall -g @noeton/logos, or delete the file), and inside Claude Code run /plugin uninstall logos@logos.