Skip to content

andrewyatesai/orca-alab

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

8,238 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Orca Orca: ALab Edition

An IDE for orchestrating fleets of AI coding agents — built on a Rust terminal engine.

Orca: ALab Edition is a desktop workspace for running CLI coding agents side by side: isolated Git worktrees, terminals, editing, an embedded browser, source control, diff review, SSH workspaces, GitHub and Linear integrations, mobile monitoring, Computer Use, and a scriptable CLI — with the terminal stack, native hot paths, and failure recovery engineered around the aterm Rust terminal engine and evidence-driven verification.

For a guided tour of the product and its engineering, see the feature walkthrough.

Downloads

Desktop builds are on the Releases page. macOS builds are ad-hoc signed for launch, not Developer ID-signed or notarized: right-click → Open on first launch. The app can check for newer macOS releases, but installation is manual from that Releases page. Windows update installation is likewise manual while ALab has no trusted publisher identity.

ALab Edition versions itself independently. Each release documents the upstream Orca version it is aligned to — the current release is aligned to upstream Orca v1.4.147.

Orca running coding agents in parallel worktrees

Why ALab Edition?

If you spend your day in terminals full of coding agents, this edition is built for you:

  • Speed. A Rust terminal engine with optimized CPU and GPU WebAssembly renderers keeps panes responsive under agent output floods, and predictive echo keeps typing instant even on slow or remote sessions.
  • Efficiency. Focus-aware rendering QoS spends your machine's power on the pane you're looking at — background terminals batch and yield instead of burning CPU.
  • Stability. Terminal sessions live in a Rust daemon with detach/reattach and session recovery, so agents keep running and their scrollback survives app restarts and crashes. Failure recovery is verified by end-to-end crash tests, not promised.
  • A terminal with personality. aterm's cursor and typing effects — trail crossfades, fresh-ink typing, feathered ribbon ends, per-session matrix rain, and the nyan-rainbow cursor default.
  • Batteries included. ALab package bundling ships compiler and solver support with the terminal, so agent workflows that build and verify code work out of the box.

The feature walkthrough shows each of these in the product; the section below covers the engineering underneath.

What ALab Edition improves

Rust and aterm terminal stack

ALab Edition replaces the xterm.js rendering and headless-terminal dependencies with a pinned aterm engine:

  • Rust terminal state, parsing, search, selection, and scrollback
  • optimized CPU and GPU WebAssembly renderers
  • a shared render-worker architecture with explicit fallback paths
  • a native Node-API terminal engine for the Electron main process
  • a Rust terminal daemon with authenticated transport, detach/reattach, and session recovery
  • Kitty keyboard handling, inline images, predictive echo, and terminal effects

The upstream aterm revision, ALab compatibility-patch digest, generated JavaScript and type bindings, both WASM binaries, byte lengths, and SHA-256 hashes are pinned together. WASM builds apply the compatibility patch in an isolated temporary checkout, so the upstream submodule stays exact and clean. The build fails if any part of that provenance drifts.

End-to-end terminal performance

The fork treats PTY ingestion, daemon transport, backpressure, parsing, rendering, and restoration as one pipeline. Its work includes bounded output queues, acknowledgement-based flow control, frame coalescing, hidden-pane resource controls, worker/GPU recovery, and reproducible latency and throughput tooling.

Performance depends on hardware and workload, so this README does not claim a universal speedup. The benchmark harnesses and regression budgets used to evaluate changes live in the repository.

Evidence-driven compatibility

Terminal and migration work is checked through several independent systems:

  • aterm-versus-xterm differential conformance corpora
  • stateful Rust-versus-Node daemon protocol parity
  • TypeScript-versus-Rust differential tests for migrated logic
  • temporal models for PTY flow control, exit delivery, and keep-tail behavior
  • explicit reliability, flake-history, renderer-size, and performance budgets

These prove specific contracts. They are not a claim that the entire application has been formally verified.

Incremental Rust migration

Selected shared logic and hot paths are moving to Rust behind parity tests, including terminal services and portions of Git, transport, parsing, and crypto logic. ALab Edition remains an Electron and React application; it is not a fully native rewrite.

Fork-safe development

The fork maintains isolated application data, daemon protocol, versioning, update feed, and telemetry policy so development does not impersonate or overwrite an upstream Orca installation. Development source launches use the visible identity Orca: ALab Edition while retaining branch and worktree metadata internally.

The build also includes pinned generated artifacts, vendored offline Rust dependencies, universal macOS native binaries, renderer chunk budgets, and strict lint, typecheck, test, and packaging gates.

Build and run from source

Prerequisites:

  • Node.js 24
  • pnpm 10
  • rustup with stable Rust 1.96 or newer
  • Xcode Command Line Tools for macOS native components

Clone with the aterm submodule and install dependencies:

git clone --depth 1 --recurse-submodules https://github.com/andrewyatesai/orca-alab.git
cd orca-alab
pnpm install --frozen-lockfile
pnpm dev

The first launch compiles the native terminal addon and Rust daemon. Later launches reuse current artifacts.

To update an existing checkout:

git pull --ff-only
git submodule update --init --recursive
pnpm install --frozen-lockfile
pnpm dev

Maintainers can advance aterm to its latest origin/main revision and regenerate the native and WASM artifacts together with:

pnpm bump:aterm
pnpm check:aterm-pin

This maintainer path additionally requires the stable wasm32-unknown-unknown target and Binaryen's wasm-opt on PATH (brew install binaryen on macOS). The bump refreshes both Cargo lockfiles as well as the native addon, Rust daemon, and committed WASM artifacts.

The bump fails closed if a downstream compatibility patch no longer applies, which forces that patch to be reviewed when upstream changes the same code.

Build and install the development CLI:

pnpm build:cli
orca-dev --help
orca-dev open
orca-dev status --json

If ~/.local/bin is not on PATH, invoke ~/.local/bin/orca-dev directly.

Validation

Run the normal contributor gates:

pnpm lint
pnpm typecheck
pnpm test
pnpm build:rust-daemon
ORCA_LOCAL_BUILD=1 pnpm build

Deeper terminal and migration checks are available separately:

pnpm parity
pnpm parity:daemon
pnpm gauntlet
pnpm spec:protocols
pnpm bench:perf
pnpm bench:check

The full test and benchmark lanes are resource-intensive. Some platform, hardware, and opt-in integration checks run only when their prerequisites are available.

Platform note

On macOS, the dedicated Computer Use helper requires user approval for Accessibility and Screen Recording before it can control other applications or capture their windows. Because ALab builds use ad-hoc signatures, macOS may ask you to approve privacy permissions again after installing a rebuilt or newer ALab build.

License and attribution

Orca: ALab Edition is distributed under the Apache License 2.0, Copyright 2026 Andrew Yates (see NOTICE).

It incorporates code from the Orca project, Copyright (c) 2026 Lovecast Inc., under the MIT License; that notice and all third-party notices are preserved in THIRD-PARTY-NOTICES.md. Report issues in this repository.

About

Orca is the next-gen IDE for working with a fleet of parallel agents. Run any coding agent with your own subscription. Available on desktop and mobile.

Resources

Contributing

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages