vocsong

VocsCode — AI skill for Claude Code

AI community

One desktop app for every coding agent.

How to install VocsCode

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

What VocsCode does

One desktop app for every coding agent. Pick the harness per session and any model it can reach.

Alternatives in AI

  • Skills Manage — Desktop app to manage AI coding agent skills across Claude Code, Cursor, Gemini CLI, Codex, and 20+ platforms 2.2k ★
  • LongHorizon Harness — The long-horizon computer-use harness 1.3k ★
  • Loushang — AI-native agent harness for coding workflows by python: multi-model LLM orchestration, stateful sessions, tool 1.3k ★

README

Vocs Code

One desktop app for every coding agent. Pick the **harness** per session and any **model** it can reach.

Harness Engine Approvals Models
Claude Agent SDK @anthropic-ai/claude-agent-sdk (Claude Code loop, hooks, MCP, checkpoints) interactive (canUseTool) Anthropic catalog, plus Bedrock/Vertex/Foundry/gateway via env
Codex (app-server) codex app-server JSON-RPC — the same engine as the Codex desktop app interactive (command + file-change requests), steer, interrupt model/list from Codex, any model_providers entry
Codex (exec SDK) @openai/codex-sdk none — sandbox mode is the boundary Codex catalog
Pi pi --mode rpc + bundled approvals extension interactive pi's registry: Anthropic, OpenAI, Codex OAuth, Google, DeepSeek, OpenRouter, Ollama, custom
ACP agent Agent Client Protocol over stdio: DeepSeek Harness (dsh --profile acp), Claude Agent ACP, Codex ACP, Pi ACP, Gemini CLI, anything else interactive (session/request_permission) agent-advertised config options
Native loop built-in loop with bash / read / write / edit / glob / grep interactive Anthropic API or any OpenAI-compatible endpoint (OpenAI, DeepSeek, OpenRouter, Ollama, LM Studio, Groq, xAI, Mistral, Gemini)

What it borrows, and what it adds

From the **Codex app**: threads grouped by project, isolated git worktrees per session, sandbox-graded permission modes, steer-vs-queue while a turn runs, diff review with per-file revert and commit, `/goal` with an iteration guard.

From the **Claude Code desktop app**: multi-session sidebar with live status, permission modes (Ask / Accept edits / Plan / Auto / Full access), live model and effort switching, cost and context tracking, a side panel with Changes / Files and a real [terminal](#terminal), slash commands and `@file` mentions, notifications when a turn needs you.

Added on top: every feature above works **