Phantom
Phantom / open engineering

Tools that make agents useful.

Phantom is the command center. The other repositories solve specific parts of the work: keeping secrets out of model context, binding the right account, turning voice into usable instructions, coding across providers, and moving a web app to native SwiftUI. Each is its own public project.

The mission: chasing open-source RSI (recursive self-improvement). Together these tools form an autonomous, code-first software factory run by agents. Agents pick work, open pull requests, merge them and ship releases through each repository's sanctioned workflow. The human owner acts as a passive board member, setting direction and stepping in only for passwords, 2FA, device approvals and payments. Read the strategy · How agents work on Phantom

—sum of repository stars across the six projects belowGitHub snapshot: loading
Public GitHub history

Follow the work.

Choose the whole set or a single repository. The line steps only when a public star was recorded.

Six repositories, one measured timeline

Cumulative GitHub repository starsUse the controls and slider below to explore dated repository star totals. ——0 ——
Loading history…

Loading GitHub history…

The chart is a dated snapshot of public GitHub star events. The aggregate adds six repo totals; it does not count unique people or measure product adoption. Open a repository for its current count.

The six selected repositories

Pick the right tool.

Independent projects with different jobs. Start with the one that solves your immediate problem.

01

Phantom

Command center

One workbench for coding-agent seats, local models, a Leader, and a fleet bounded by signed authority.

—stars
02

Phantom Secrets

Secret boundary

Keep real secrets out of agent prompts by giving tools local, replaceable handles instead.

—stars
03

Locus

Account identity

Bind agent tools to the intended account and tenant before a command can act.

—stars
04

Lexicon

Voice vocabulary

Correct the names and technical words speech recognition gets wrong before they reach an agent.

—stars
05

AshlrCode

Coding agent CLI

Work in the terminal with a coding agent that can use tools, skills, and multiple model providers.

—stars
06

Morphkit

Web to SwiftUI

Translate a TypeScript or React app into a native SwiftUI starting point for iOS.

—stars
More agent tooling

Companions for agent sessions.

Also independent and installed separately. They are not in the star chart above; open each repository for its current state.

07

Ashlr Plugin

Compact tool results

Give coding agents compact tool results and keep a local usage ledger.

08

Ashlr MD

Markdown

Edit Markdown through CLI and MCP interfaces.

09

Ashlr Stack

Service setup

Configure supported development services through CLI and MCP.

Phantom can reach companions through its MCP gateway; registration alone does not prove a tool is ready or authorized. See the companion setup guide.

Start building in Phantom.

Install the CLI, open the console, and follow the guide for your first seat.

Install and get started

Pick a use-case guide.

Own accounts, local models, signed autonomy and review workflows, each claim linked to its source.

Browse the guides

Make a project better.

Each repository has its own source, issues, and contribution path. Pick one that matters to you. For Phantom, contribute with an agent from a fork.

Browse all public repositories