ghostwright

ghostwright / phantom

Public

An AI co-worker with its own computer. Self-evolving, persistent memory, MCP server, secure credential collection, email identity. Built on the Claude Agent SDK.

ghostwright.devphantom ai-agents ai-coworker anthropic autonomous-agents bun
23
0
100% credibility
Found Mar 30, 2026 at 23 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

Phantom is an AI co-worker that runs on its own dedicated virtual machine, enabling it to install software, manage databases, build dashboards, and retain context across sessions.

Star Growth

See how this repo grew from 23 to 23 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is phantom?

Phantom is a TypeScript AI agent acting as your co-worker on a dedicated Docker VM, chatting via Slack, Telegram, email, or webhooks. It solves ephemeral AI woes by persisting memory across sessions, self-evolving configs, and building shareable dashboards, APIs, or automations on its own public domain. This phantom github repository delivers a real agent built on Claude Agent SDK—no more localhost-only outputs or forgotten context.

Why is it gaining traction?

Unlike github phantom wallet clones or phantom github minecraft tools, Phantom gives the AI full autonomy: it installs databases like ClickHouse, creates MCP tools for reuse with Claude Code, and even adds channels like Discord dynamically. Developers hook on the "own computer" model—24/7 availability, encrypted secrets, scheduled jobs—yielding production results like analytics platforms from casual asks.

Who should use this?

Software engineers automating PR summaries or spinning dev Postgres; sales reps building prospect profiles; data analysts loading datasets into queryable stacks. Perfect for backend devs tired of manual ETL, ops teams monitoring infra, or solo founders needing a persistent agent for reports and prototypes.

Verdict

Deploy via Docker Compose for a capable, evolving co-worker—strong docs, 770 passing tests, Apache 2.0. With 23 stars and 1.0% credibility, it's raw but promising; polish the quickstart and it'll fly.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.