ctx

ctx is a local-first context manager for Claude Code and Codex that solves transcript drift and context loss between AI coding sessions. It binds each workstream to the exact Claude and/or Codex conversation it originated from, so later pulls stay on the correct conversation instead of jumping to the newest chat on disk. Everything runs locally with no API keys required — data is stored in plain SQLite plus local files.

• Exact transcript binding: each internal ctx session binds to the precise Claude and/or Codex conversation it came from, preventing drift across sessions.
• Safe branching: use ctx branch source-stream target-stream to create a new workstream seeded from the current saved state of another without sharing future transcript pulls or hijacking the source conversation.
• Workstream management: start, resume, rename, delete, and list workstreams via /ctx slash commands in Claude Code or ctx subcommands in Codex.
• Curated memory loads: pin saved entries so they always load, exclude entries to keep them searchable without passing them back to the model, or delete them entirely via the ctx curate terminal UI.
• Indexed retrieval: saved workstreams, sessions, and entries are indexed for fast ctx search lookup across all stored context.
• Local browser frontend: run ctx web --open to browse, search, and copy continuation commands from an optional local browser UI.
• Multiple install paths: clone and run ./setup.sh for a project-local setup, use ./setup.sh --global for a shared global install, or bootstrap via curl for agent shell environments.
• Multi-repo support: set ctx_DB to a shared path so multiple repos can share the same context database.
• No external dependencies: no API keys, no hosted service — plain SQLite and local files only, with MIT license.