AI Skill Hub 推荐使用:Pi-Shell ACP桥接 是一款优质的MCP工具。AI 综合评分 7.2 分,在同类工具中表现稳健。如果你正在寻找可靠的MCP工具解决方案,这是一个值得深入了解的选择。
Pi-Shell ACP桥接 是一款遵循 MCP(Model Context Protocol)标准协议的 AI 工具扩展。通过 MCP 协议,它可以让 Claude、Cursor 等主流 AI 客户端直接访问和操作外部工具、数据源和服务,实现 AI 能力的无缝扩展。无论是文件操作、数据库查询还是 API 调用,都可以通过自然语言在 AI 对话中直接触发,极大提升生产效率。
Pi-Shell ACP桥接 是一款遵循 MCP(Model Context Protocol)标准协议的 AI 工具扩展。通过 MCP 协议,它可以让 Claude、Cursor 等主流 AI 客户端直接访问和操作外部工具、数据源和服务,实现 AI 能力的无缝扩展。无论是文件操作、数据库查询还是 API 调用,都可以通过自然语言在 AI 对话中直接触发,极大提升生产效率。
# 方式一:通过 Claude Code CLI 一键安装
claude skill install https://github.com/junghan0611/pi-shell-acp
# 方式二:手动配置 claude_desktop_config.json
{
"mcpServers": {
"pi-shell-acp--": {
"command": "npx",
"args": ["-y", "pi-shell-acp"]
}
}
}
# 配置文件位置
# macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
# Windows: %APPDATA%/Claude/claude_desktop_config.json
# 安装后在 Claude 对话中直接使用 # 示例: 用户: 请帮我用 Pi-Shell ACP桥接 执行以下任务... Claude: [自动调用 Pi-Shell ACP桥接 MCP 工具处理请求] # 查看可用工具列表 # 在 Claude 中输入:"列出所有可用的 MCP 工具"
// claude_desktop_config.json 配置示例
{
"mcpServers": {
"pi-shell_acp__": {
"command": "npx",
"args": ["-y", "pi-shell-acp"],
"env": {
// "API_KEY": "your-api-key-here"
}
}
}
}
// 保存后重启 Claude Desktop 生效
Use Claude Code, Codex, and Gemini CLI through Agent Client Protocol (ACP) inside pi.

· maintained by junghanacs.com
Public, active development. Real working code, still young. Verify it in your own workflow before relying on it all day. Evidence calibration: VERIFY.md.

pi
→ pi-shell-acp
→ claude-agent-acp | codex-acp | gemini --acp
→ Claude Code | Codex | Gemini CLI
pi-shell-acp is a thin ACP provider for pi: no OAuth proxy, no CLI transcript scraping, no Claude Code emulation. It connects pi to locally authenticated ACP backends with no core patch and no bypass. Each backend keeps its own model, API, and tool semantics; the bridge shapes only the pi-facing operating surface.
Direction. Inverse ofpi-acp.pi-acplets external ACP clients talk to pi;pi-shell-acplets pi talk to ACP backends.
Anthropic subscription billing. From 2026-06-15, Anthropic third-party agent paths (ACP, Agent SDK,claude -p, pi-shell-acp's Claude backend) consume a separate Agent SDK credit pool, distinct from Claude chat and theclaudeCLI used as an interactive terminal.pi-shell-acprespects that distinction — no bypass, no emulation — and preserves capability dignity across all three backends (see AGENTS.md invariants #7, #9, #10). The recommended default runtime leans toward paths outside Anthropic's Agent SDK metering (Codex / Gemini); Claude remains a strong coding worker invoked when its quality is worth the credit cost. The operator decides the mix.
Claude / Codex ACP server packages (@agentclientprotocol/claude-agent-acp, @zed-industries/codex-acp) ship as pinned dependencies of pi-shell-acp; backend authentication still belongs to the operator's local CLI / runtime. Once the bridge is installed, the resolver picks the ACP server in this order:
CLAUDE_AGENT_ACP_COMMAND / CODEX_ACP_COMMAND env override — explicit override for an alternative binary or a wrapper command.require.resolve(...) against the bundled package dependency — @agentclientprotocol/claude-agent-acp for Claude, @zed-industries/codex-acp for Codex. This is the default path; no extra global install needed.PATH:claude-agent-acp / PATH:codex-acp fallback — used when the package resolution fails (e.g. a hand-edited node_modules).Codex smoke (no global install required — the codex-acp pinned in dependencies is resolved automatically):
./run.sh smoke-codex /path/to/your-project
To force a global codex-acp (PATH fallback or development override):
pnpm add -g @zed-industries/codex-acp@0.14.0
Gemini is different — the gemini CLI binary is itself the ACP server, not a separate *-acp server package. It must be installed and authenticated on the operator's machine. Curated model: pi-shell-acp/gemini-3.1-pro-preview.
pnpm add -g @google/gemini-cli
gemini # one-time interactive login (oauth-personal) or set GEMINI_API_KEY
./run.sh smoke-gemini /path/to/your-project
Backend is inferred from the model — Anthropic → claude, OpenAI → codex, Gemini → gemini. Set backend only to pin.
pi-shell-acp is a thin ACP bridge — it connects pi to a local Claude/Codex/Gemini backend the operator has already installed and authenticated. The bridge does not provide Claude credentials, tokens, or subscription access, and does not bypass any backend auth. Whatever the operator's local claude / codex / gemini already trusts is what pi-shell-acp uses.
pi accepts four install sources for the bridge — npm: or git:, each in global (default, writes to ~/.pi/agent/settings.json) or project (-l flag, writes to .pi/settings.json) scope. A fifth path is a local clone for hacking on the bridge.
After installing the package, run run.sh install . in your target project. The script writes the piShellAcpProvider block into .pi/settings.json with the correct absolute path for pi-tools-bridge/start.sh — no hand-editing required. The exact location of run.sh depends on which install path was used (each section below shows it). For manual configuration, pi/settings.reference.json is the reference shape — see Settings below.
Reference shape lives in pi/settings.reference.json. Minimum:
{
"compaction": { "enabled": false },
"piShellAcpProvider": {
"appendSystemPrompt": false,
"settingSources": [],
"strictMcpConfig": true,
"showToolNotifications": true,
"tools": ["Read", "Bash", "Edit", "Write"],
"skillPlugins": [],
"permissionAllow": ["Read(*)", "Bash(*)", "Edit(*)", "Write(*)", "mcp__*"],
"mcpServers": {
"pi-tools-bridge": {
"command": "/path/to/pi-shell-acp/mcp/pi-tools-bridge/start.sh",
"args": []
}
}
}
}
mcpServers is the only ACP MCP injection path. In practice this repo is about the bundled pi-tools-bridge, which carries pi capabilities into ACP-backed sessions — not about being a general MCP catalog. Invalid entries throw McpServerConfigError — broken tool state surfaces as broken tool state. ./run.sh install writes the bundled pi-tools-bridge entry and prunes the legacy bundled session-bridge entry from older installs.
appendSystemPrompt: false is intentional. Pi / AGENTS context rides the first-user augment; putting it into the Claude _meta.systemPrompt carrier can route OAuth sessions to metered "extra usage" billing.
resume > load > new idea originpi install git:github.com/junghan0611/pi-shell-acp
cd /path/to/your-project
~/.pi/agent/git/github.com/junghan0611/pi-shell-acp/run.sh install .
~/.pi/agent/git/github.com/junghan0611/pi-shell-acp/run.sh smoke-all .
创意性的跨协议桥接方案,整合多个主流AI模型。但项目早期阶段,成熟度���社区支持度有限,适合尝鲜和定制化集成场景。
AI Skill Hub 为第三方内容聚合平台,本页面信息基于公开数据整理,不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后,再部署至生产环境,并做好必要的安全评估。
✅ MIT 协议 — 最宽松的开源协议之一,可自由商用、修改、分发,仅需保留版权声明。
总体来看,Pi-Shell ACP桥接 是一款质量良好的MCP工具,在同类工具中具备一定竞争力。AI Skill Hub 将持续追踪其更新动态,建议收藏备用,结合自身场景选择合适时机引入使用。
| 原始名称 | pi-shell-acp |
| 原始描述 | 开源MCP工具:ACP bridge provider for pi — use Claude Code, Codex, and Gemini via official ACP。⭐6 · TypeScript |
| Topics | MCPACP协议多模型集成AI代理TypeScript |
| GitHub | https://github.com/junghan0611/pi-shell-acp |
| License | MIT |
| 语言 | TypeScript |
收录时间:2026-05-22 · 更新时间:2026-05-22 · License:MIT · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型,复制安装指令后粘贴到对应客户端