Claude Code lives entirely in the terminal, which is fine until you juggle ten projects, a dozen half-finished sessions, and a token bill nobody can explain. The useful idea here is that opcode never replaces that terminal workflow — it reads Claude Code's own files under ~/.claude/projects/ and renders them as a visual cockpit, so updates from Anthropic don't break it.
What Sets It Apart
- It is a view layer, not a fork. Sessions, history, and config come straight from Claude Code's local files, so resuming a session in opcode is the same session you started in the CLI.
- Custom agents are reusable definitions — a system prompt plus scoped permissions — that run in isolated processes. The payoff is sandboxing: you can hand an agent a narrow slice of access instead of granting blanket file and shell rights.
- Checkpoints give git-like branching of a conversation. Roll a session back to an earlier state and try a different prompt path without losing the original thread.
- Built on Tauri 2 (Rust backend, React frontend): a small binary, local-only storage, and no telemetry, licensed AGPL-3.0.
A Note on the Name
The project shipped first as "Claudia" and was renamed to opcode after the name collided with Anthropic's Claude trademark. It is built by the team at Asterisk and is explicitly independent — not affiliated with, endorsed by, or sponsored by Anthropic.
Who It's For
Great fit if you run several Claude Code projects in parallel and want session resume, per-project cost visibility, agent reuse, and MCP server management in one window. Look elsewhere if you prefer a pure keyboard-driven terminal flow, need official vendor support, or are on a platform without a maintained desktop build — opcode is a community project distributed as desktop binaries.
