Averlo for macOS · Private Beta

A spatial workspace for autonomous coding agents.

Run Claude Code, test suites, and local browsers side by side on an infinite canvas. Interconnected through the Model Context Protocol, shared memory, and high-frequency voice.

Claim Founder pass ($149)
macOS 14.0+ Claude 3.5 Sonnet native Model Context Protocol (MCP)
averlo — main workspace (3 active agents)
mcp :8766 connected
Claude Code Working
$ claude refactor src/auth.ts
Addressing token rotation logic...
→ Delegating tests to @codex
bruno_prompt_agent("codex", "run auth")
Codex Testing
$ npm test -- auth.spec.ts
✓ 14 suites passed (42ms)
Reporting to @claude via MCP
WebKit Preview :3000
localhost:3000/login
Hot reload verified
Voice: “Ask Claude to refactor auth and have Codex run tests”
0FPS
Native Swift canvas, no dropped frames
0ms
Last full auth suite, green on canvas
0agents
Claude, Codex, WebKit — one workspace
$0
Paid once. Yours across v1.x and v2.x
The Architecture
Single chat threads cannot represent the complexity of multi-agent software engineering.

When software developers write code today, they don't do it in isolation. A modern task requires writing implementation code, verifying tests in a secondary process, inspecting terminal logs, and checking browser renders simultaneously.

Averlo provides an infinite, hardware-accelerated spatial canvas where multiple instances of Claude Code, test runners, and browsers live side by side.

Agents communicate directly over the Model Context Protocol (MCP). Claude can autonomously inspect what Codex is doing, delegate sub-tasks, and receive verification without human copy-pasting.

Key Capabilities
01 / CANVASES

Infinite 120 FPS Canvas

Built with native Swift and SwiftTerm. Arrange, pan, and zoom across dozens of agent cards without losing terminal scrollback or context.

02 / PROTOCOL

Model Context Protocol

Every card exposes a standardized MCP bridge. Claude Code can spawn sibling agents, query their status, and send instructions programmatically.

03 / VOICE

Continuous Speech VAD

Hands-free voice recognition designed for full compound instructions. It never cuts you off mid-sentence and routes your prompt directly to the target agent.

Pricing

One clean lifetime license.

No monthly subscription anxiety. Support early independent software development.

Early Preview

For engineers wanting to test early macOS builds and give feedback.
$0 Free during public preview
  • Full macOS spatial canvas
  • Connect Claude Code and local CLIs
  • Local MCP server runtime
  • Community updates and support
Limited to 100 builders

The canvas is quiet.
Bring your agents.

Founder passes include every v1.x and v2.x release, commercial use, and a direct line to the person building Averlo.