★ 713

/

Meet Shelley

Installation and Configuration A Note on Shelley’s Security Model Mac Linux Phone exe.dev From Source Connecting to Models

Features Audio and Screen Cast Input Mobile and desktop apps Automatic Compaction Customization Hooks Diffs and Comments and Diff Highlights Embedded Terminal output_iframe Images and Audio Self Inspection (and sqlite) Subagents, Asynchronous Tools Multiple Simultaneous Conversations Relatively Lightweight State of the Art Browser Tools Profiles

Essays and Blogs

Shelley's Values and Principles 1. Shelley must be ready when inspiration strikes 2. Multi-Modal in both input and output 3. Dev tools must be open source and customizable 4. Transparent 5. Self-introspective

Written in Go Why Go The web UI is Vue Will Shelley always be written in Go?

Blog Highlights exe.dev blog sketch.dev blog

Contributing

Development Architecture Testing What to Look for in Reviews Invariants We Care About The “Demo Server”

FAQ Why is it called Shelley? For exe.dev VMs, how Is the Shelley icon chosen? How is Shelley different from other coding agents? I already use Claude Code. Why would I use Shelley? Is Shelley meant for serious work, or just small projects? Can I run Shelley outside exe.dev? How do I put a password on Shelley? Can I bring my own API key, or a local model? Can I use my Claude or ChatGPT subscription? Which models work best with Shelley? Does Shelley read AGENTS.md or CLAUDE.md? Can I see or change Shelley’s system prompt? Does Shelley support skills and MCP? Is there a plan mode? Can Shelley hand work to Claude Code or Codex? Can I close the tab while Shelley is working? Can I use Shelley from a script or the command line? How do I update Shelley? Where are my conversations stored, and can I export them? How do I keep Shelley away from my secrets? Do you take AI-written pull requests?

Essays and Blogs

Shelley's Values and Principles
Written in Go
Blog Highlights

← Features Shelley's Values and Principles →

made by exe.dev