About

PHEWSH is not another chatbot, not another AI agent framework, and not a wrapper around any single model.

It is a continuity layer: portable project context, intent synchronization, and cross-tool identity so your AI working state persists regardless of which model, runtime, or tool is active today.

Claude Code changes? Fine. Switch to Cursor? Fine. A new model drops tomorrow? Fine. Your project context, intent artifacts, and working memory follow you — not the tool you're using.

What that means in practice

Define your project once — using the 12-node intent compass in the web app, or by writing a .intent/ directory by hand. That spec is your portable project document: your vision, plan, next actions — nothing ambiguous, nothing locked to a single vendor.

The sequencer is an emerging distribution layer that emits the correct context format for each environment from a shared .intent/ source — CLAUDE.md for Claude Code, .cursorrules for Cursor, MCP JSON for agents. One spec. Every tool speaks it.

.intent/
the sequencer emits ↓
CLAUDE.md .cursorrules MCP JSON phewsh CLI
One spec. Every tool speaks it.

Sign into PHEWSH once and your project state is saved. Switch machines — your context is there. Re-open the web app — your artifacts are there.

Why login matters

Without an account your context lives only in the browser tab where you created it, or in a .intent/ folder that doesn't update when you switch machines. Login turns local intent into persistent continuity.

The mental model is simple: "Sign in once, your AI context follows you everywhere."

How it differs from Honcho

Honcho focuses on runtime and session orchestration. PHEWSH owns the portable intent layer and the continuity fabric that lives on top of whatever runtime is active. Run phewsh seq --target json and point Honcho at the output — cooperation, not competition.