roowus

Truss — AI skill for Claude Code

AI community

One web UI for all your AI agent harnesses (pi, Claude Code, DeepSeek Harness, Hermes).

How to install Truss

This entry records only its repository, not the path inside it, so there is no exact command to give. Open roowus/truss and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Truss does

One web UI for all your AI agent harnesses (pi, Claude Code, DeepSeek Harness, Hermes). Dockable tabs, real terminals, an LLM trajectory view, todos and an inbox your agents file work into. Very beta, self-hosted, single-user.

Alternatives in AI

Desktop Cc Gui — Multi-engine AI coding desktop client (Tauri) 4.4k ★
EchoBird — One-click install + model switch:Claude Code,Codex CLI (OpenAI), Grok Build (xAI), DeepSeek Harness, Kimi Code 3.1k ★
TokenTracker — Local-first AI token usage & cost tracker for 31 coding tools incl 1.4k ★

README

Truss

**The head for your harnesses.**

Truss is a web frontend for AI agent harnesses. It is not a harness itself, it hosts them. Run [pi](https://www.npmjs.com/package/@earendil-works/pi-coding-agent), [Claude Code](https://www.npmjs.com/package/@anthropic-ai/claude-code), [DeepSeek Harness](https://www.npmjs.com/package/@deepseek-ai/dsh), and Hermes side by side, pick a harness per chat, and watch all of them through the same interface. You get dockable tab windows, real terminals, a trajectory view that works like the Chrome network tab but for model calls, context tracking, a subagent tree, a task board, todos, and an inbox for things agents hand back to you.

![Truss workspace](docs/screenshots/workspace.png)

[!WARNING] **Truss is very beta.** One person builds it and it moves fast. Expect breaking changes between commits, rough edges, and migrations that might not be graceful. **Do not put it on the public internet** (see [Security model](#security-model)). Back up `apps/server/data/` if you care about what's in it. That said, it is a real working app that gets used every day, and issues and PRs are welcome.

What you get

Every harness in one UI. Chat, streaming, interrupt, permission cards, and dead sessions that resume with their history intact.
A real workspace. Split panes any direction, each pane has its own tab bar, tabs drag between panes and between workspaces (think virtual desktops), and the layout persists. Tabs shrink to fit like Chrome; the close button shows on hover, or sits inline when there's room.
Trajectory. Every LLM call with model, latency, tokens, cost (cache reads and writes too). Expand a call to see the tools that ran inside it. Retries and failures stand out.
Feed. Your inbox. Permission requests you can answer from the card, finished-work notes, crashes, context-window warnings, and reports agents post. Sort, filter, save, dismiss, or share a card into another agent's chat.
Todos.