Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.tryscott.ai/llms.txt

Use this file to discover all available pages before exploring further.

The scott CLI is a single self-contained binary (~50MB). No Node, no Homebrew, no npm required.

​
Install

curl -fsSL https://tryscott.ai/install | sh
Installs the binary at /usr/local/bin/scott (or ~/.scott/bin/scott if /usr/local/bin isn’t writable). macOS and Linux supported; Windows coming.

​
Commands

​
scott login

Authenticates the CLI via your browser. Opens app.tryscott.ai/auth/cli, then receives the session token over a local loopback so it never leaves your machine via any third party. 
scott login
Token is stored at ~/.scott/credentials.json (mode 0600). Run once per machine.

​
scott setup

Wires Scott’s skills, rule, and hosted MCP server entry into your coding agent. Run after scott login. 
scott setup --claude        # Set up for Claude Code
scott setup --cursor        # Set up for Cursor
scott setup --codex         # Set up for Codex
scott setup --opencode      # Set up for OpenCode
scott setup --gemini        # Set up for Gemini CLI
Multiple agents in one run: 
scott setup --claude --cursor
By default this writes all three layers — skills + rule + MCP server entry. Flags:
  • --no-mcp — skip the hosted MCP server entry (write only skills + rule)
  • --mcp-only — write only the MCP server entry (skip skills + rule)
  • --project — configure for the current project instead of globally

​
scott push

Captures the active Claude Code session in the current directory and creates a planspace from it. Zero flags — the command finds the session, parses the transcript with full attribution, and posts to Scott. 
cd your-project              # in a Claude Code project
scott push
The CLI prints the planspace URL on its own line for command-click. The auto-summary turn kicks off in the background; the planspace UI shows live progress. In practice you rarely run scott push directly. Once scott setup is done, ask your coding agent to “push this to Scott” — the skill loaded by scott setup makes the agent run scott push for you.

​
scott remove

Inverse of scott setup. Removes Scott’s skills, rule, and MCP entry from the chosen agent without touching the rest of its config. 
scott remove --claude
Same --no-mcp / --mcp-only / --project flags as setup.

​
scott logout

Clears the stored credentials at ~/.scott/credentials.json. 
scott logout

​
Typical workflow

# One-time setup
curl -fsSL https://tryscott.ai/install | sh
scott login
scott setup --claude

# Day-to-day: just use Claude Code normally.
cd ~/some-project
# (open Claude Code, have a conversation)

# When you're ready to share or persist the work:
# Ask the agent — "push this conversation to scott"
# The agent runs scott push for you.
# A planspace URL appears in the terminal. Command-click → planspace
# opens in the browser with the conversation summary.

​
FAQ

No. The CLI is a single self-contained binary. No Node, no Python, no Homebrew, no npm. macOS and Linux are supported on both Apple Silicon and Intel.
Claude Code today. Codex is wired into the skills layer (scott setup --codex) but transcript parsing for Codex is on the roadmap.
Yes. The CLI works on its own — it talks to Scott over HTTPS. After scott push, open the printed URL in either the web app or the desktop app.
Per agent:
  • A rule file (e.g. .claude/rules/scott.md for Claude Code)
  • Two skills (scott-push and scott-fetch) so the agent knows when to call the CLI vs the MCP tools
  • One MCP server entry pointing at https://mcp.tryscott.ai/mcp (skip with --no-mcp)
Run scott remove --<agent> to reverse it cleanly.
~/.scott/credentials.json (mode 0600, parent dir 0700). Cleared by scott logout.