Phantom markets AI coworker with own computer
Phantom is an open-source AI coworker built on the Claude Agent SDK, with persistent memory, MCP support, secure credential collection, and an email identity. It is aimed at long-running autonomous work on a dedicated machine rather than one-off chat sessions.
This is less “AI assistant” and more “persistent operator” branding, and that distinction matters: Phantom is trying to own the category where agents keep state, build tools, and stay online across sessions.
- –Persistent memory and its own VM make it viable for work that needs continuity, not just prompt-response loops
- –MCP support turns it into a tool-using hub, which is the right abstraction for autonomous agents that need to extend themselves
- –Secure credential collection plus Slack/email identity push it toward real operational use, but also raise the bar on trust and safety
- –The strongest pitch is long-running work: monitoring, analytics, and automations that improve over time instead of resetting every run
- –As an open-source release, it is more compelling as an architecture reference than as a polished end-user product
DISCOVERED
59d ago
2026-03-31
PUBLISHED
59d ago
2026-03-31
RELEVANCE
AUTHOR
Github Awesome
