- Author
- Not stated
- Published
- Undated
- Kind
- Project documentation file describing a self-hosted tool
- Original
- github.com/lastnpcalex/a-shadow-loom/blob/braid/LOOM_EXPLAINER.md
- Subject
- Harnesses and agent loops
What it is
The file is an overview written by the project for its own tool. It describes the Loom as a multi-engine orchestration layer. It names the engines it can drive and the features around them.
What it says
- The document says no single model or agent design is best for every task.
- The Loom is a shared layer, and model clients are interchangeable shuttles.
- A web interface over a Tailscale tailnet lets users control agents from a phone.
- Progressive draft saving keeps a run going if the browser tab closes.
- A permission hook halts agent tool calls and asks the user in the browser.
- One database schema lets users switch models mid-conversation. A tree view shows every branch.
What evidence it gives
The document only describes features and links to source files. It shows no demos, tests or measurements.
Interests and limits
The bookmarked path returns a not-found error. The file was read from docs/architecture/LOOM_EXPLAINER.md on the braid branch instead. The authors wrote it about their own tool. No claim was checked against the code. No author or date is stated.