Arcane RMCP
ai.dinglebear由 dinglebear.ai 发布v0.4.4更新于 Oct 24, 2026
Arcane Docker and Compose management over MCP and CLI with authenticated stdio and HTTP.
发布方自己的服务:RMCP
概览
一个本地 MCP 桥接服务,让助手通过已有的 Arcane API 服务器查看和管理 Docker 与 Compose 资源。
- 可以做什么
- 通过单个 arcane 工具(配合 action 与 subaction 参数)或对应的 CLI 命令,读取 Arcane 状态以及 Docker 环境、项目、容器、镜像、网络、卷、注册表、GitOps、更新、漏洞和系统状态。创建、更新、启动、停止、重启、删除、清理、部署、同步、扫描并备份受支持的 Arcane 资源。检查镜像更新并查看漏洞发现,包括忽略或取消忽略。提供 quick_start 提示和 schema 资源,供客户端发现用法。
- 适用场景
- 当你已经运行 Arcane 服务器,并希望助手从本地 MCP 客户端查询或操作其 Docker 环境和 Compose 项目,而不必在 Arcane 界面中手动点击或手写 API 调用时使用。
- 运行要求
- 以本地进程运行,可通过 npx(npm 包会从 GitHub Releases 下载匹配的二进制文件)、发布安装脚本、Docker 或源码构建安装。需要一个已有的 Arcane API 服务器、RARCANE_API_URL 基础地址和 RARCANE_API_KEY 密钥,通过环境变量或配置文件提供。需要访问 Arcane 服务器和 GitHub Releases 的网络。
安装
在 SourceWeft 中
- 打开 控制台中的 Arcane RMCP,将其添加到工作区。
- 为需要使用其工具的对话启用该服务。
Desktop only,通过 STDIO。 STDIO 服务会启动本地进程,因此需要 SourceWeft 桌面宿主。
其他 MCP 客户端
参照 仓库 中的启动说明。
README
arcane-rmcp
Arcane Docker and Compose management over MCP and CLI with authenticated stdio and HTTP.
It exposes one MCP tool, arcane, plus the rarcane CLI. Agents can inspect
Arcane environments, manage compose projects, containers, images, networks,
volumes, registries, GitOps syncs, image updates, vulnerability findings, and
system operations through stdio MCP, Streamable HTTP MCP, or direct shell
commands.
30-second path: set RARCANE_API_URL and RARCANE_API_KEY, then run
npx -y @dinglebear/rarcane status -> start loopback HTTP with
RARCANE_MCP_HOST=127.0.0.1 npx -y @dinglebear/rarcane serve -> call tools/call with
{"action":"status"}.
Status: operational RMCP upstream-client server. Write-capable; destructive
Docker and Arcane operations require explicit confirmation. HTTP MCP supports
loopback dev mode, static bearer tokens, and Google OAuth through lab-auth.
Not for: replacing Arcane, bypassing Docker or Arcane authorization, running arbitrary shell commands, storing registry or Git credentials, multi-tenant isolation, or passing Arcane API keys through MCP tool arguments.
Contents
- Naming
- Capabilities And Boundaries
- Install
- Quickstart
- Client Configuration
- Runtime Surfaces
- MCP Tool Reference
- CLI Reference
- Configuration
- Authentication
- Safety And Trust Model
- Architecture
- Distribution Contract
- Development
- Verification
- Deployment
- Troubleshooting
- Related Servers
- Documentation
- License
Naming
The repo and npm package use the upstream service name, while the shipped
binary keeps the historical Rust CLI name rarcane. The MCP server may be
registered as rarcane, but the tool clients call is arcane.
Capabilities And Boundaries
- Read Arcane status plus Docker environment, project, container, image, network, volume, registry, GitOps, update, vulnerability, and system state.
- Create, update, start, stop, restart, delete, prune, deploy, sync, scan, and back up supported Arcane resources through action/subaction dispatch.
- Enforce action scopes and destructive-operation confirmation before forwarding write operations to Arcane.
- Expose the
quick_startprompt andrarcane://schema/mcp-toolresource for client-side discovery. - Provide setup, doctor, and watch commands for local plugin/runtime checks.
Install
npm / npx
Run the stdio MCP server or CLI without a manual binary install:
The npm package downloads rarcane during postinstall. Override download
behavior only when testing packaging:
Build From Source
Minimum supported Rust version: 1.90.
Quickstart
1. Configure Arcane
Point the bridge at an existing Arcane API server:
The API key is read from env or config only. Do not pass it in MCP tool
arguments or CLI --params-json.
2. Run A Safe CLI Call
3. Start Loopback HTTP MCP
In another shell:
4. Make A First MCP Call
Client Configuration
Claude Code Stdio
Claude Code HTTP
Codex / Labby Gateway
Register Arcane through Labby as an HTTP upstream when sharing one long-running server, or run it directly as stdio for local-only use.
Generic MCP JSON
Do not put RARCANE_API_KEY, registry credentials, Git credentials, OAuth
secrets, SSH keys, passwords, or upstream bearer tokens in MCP tool arguments.
Use env, config files, or the MCP client's secret storage.
MCP callers never provide credentials, tokens, keys, or secrets as action
arguments.
Runtime Surfaces
MCP Tool Reference
One MCP tool is exposed: arcane. Pass the required action argument and, for
most domains, a required subaction.
Action specs in src/actions.rs and the generated schema notes in
docs/MCP_SCHEMA.md are the source of truth for required parameters.
Example read-only tool arguments:
Example destructive operation:
CLI Reference
rarcane exposes the same service layer as the MCP tool:
--params-json accepts action-specific JSON payloads. Do not use it to pass
credentials.
Configuration
Host installs read ~/.rarcane/.env, ~/.rarcane/config.toml, and process env.
Containers read /data/.env, /data/config.toml, and process env.
Authentication
Stdio MCP runs as a local trusted child process and does not use HTTP auth.
HTTP MCP auth policy:
OAuth mode uses the RARCANE_MCP_* Google OAuth variables documented in
docs/CONFIG.md.
OAuth mode is single-replica: client/session state and signing keys are local to one host. Do not run multiple OAuth replicas until those stores are externalized or safely shared.
Safety And Trust Model
- Arcane API credentials are loaded from config/env only.
- MCP callers select actions and payloads, not upstream credentials.
- Destructive subactions reject unless
params.confirm=trueor CLI--confirmis present. - Unknown actions, unknown subactions, missing environment IDs, missing IDs, and unsafe volume paths are rejected before upstream calls.
- Non-loopback HTTP deployments must use bearer auth, OAuth, or a trusted authenticated gateway.
- This bridge does not sandbox Arcane itself. Arcane remains responsible for Docker authorization and the effects of Docker operations.
Architecture
Distribution Contract
Cargo.toml,Cargo.lock,packages/arcane-rmcp/package.json,.release-please-manifest.json, andserver.jsonmust agree on the released version.- GitHub Releases publish the
rarcanebinary consumed by the npm launcher. - The npm package name is
@dinglebear/rarcane; the installed binary alias israrcane. - Docker/OCI metadata uses
ghcr.io/dinglebear-ai/rarcane:<version>. plugins/rarcane/.mcp.jsonmust launchnpx -y @dinglebear/rarcane mcpso stdio clients start the MCP transport rather than the HTTP server.- The root README is curated. Generated or source-of-truth details live in
src/actions.rs,docs/MCP_SCHEMA.md, and the package/registry manifests.
Development
Verification
Runtime smoke:
HTTP smoke:
Deployment
Use loopback for local development:
Use Docker Compose for shared HTTP deployment:
When binding to a non-loopback address, configure RARCANE_MCP_TOKEN,
RARCANE_MCP_AUTH_MODE=oauth, or RARCANE_NOAUTH=true behind an authenticated
gateway.
Troubleshooting
Related Servers
- soma - RMCP runtime for provider-backed MCP servers.
- unifi-rmcp - UniFi controller REST API bridge.
- tailscale-rmcp - Tailscale API bridge for devices, users, and tailnet operations.
- unraid - Unraid GraphQL bridge for NAS and server management.
- apprise-rmcp - Apprise notification fan-out bridge for many delivery backends.
- gotify-rmcp - Gotify push notification bridge for sends, messages, apps, and clients.
- yarr - Media-stack bridge for Sonarr, Radarr, Prowlarr, Plex, and related services.
- ytdl-rmcp - Media download and metadata workflow server.
- synapse-rmcp - Local Synapse workflow server for scout and flux actions.
- cortex - Syslog and homelab log aggregation MCP server.
- axon - RAG, crawl, scrape, extract, and semantic search project.
- labby - Homelab control plane and MCP gateway project.
- lumen - Local semantic code search MCP server.
Documentation
docs/API.mdis the curated action-contract overview.docs/CONFIG.mdis the curated configuration and auth reference.docs/QUICKSTART.mdis the curated smoke-test guide.docs/MCP_SCHEMA.mdis the generated/schema-drift contract for actions, resources, prompts, and validation rules.plugins/rarcane/skills/rarcane/SKILL.mdis the agent usage guide.
License
Original Dinglebear-authored portions of this project are licensed under AGPL-3.0-only. Separate commercial licensing is available for organizations that need terms outside the AGPL. Third-party material remains under its original license. See LICENSING.md.
来源:README.md,提交 7d16337
工具
0版本历史
1- v0.4.4最新Sep 16, 2026
评分
已满足适用检查项的 50%
该评分衡量此收录条目的完整程度,依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价,也不表示服务器安全或当前可用:可连通性是带日期的检查结果,评分也不包含安装量或用户评价。
- 近 30 天内可连通25 分
请保持端点能够响应 MCP 握手;SourceWeft 每 30 天探测一次。
- 已声明认证方式10 分
请在注册中心条目中声明凭据请求头,或保持端点能够响应探测,以便判定其认证方式。
- 已列出工具15 分
请让服务器在无凭据时响应 tools/list,或发布可在沙箱中无凭据启动的软件包。
- 描述必需15 分
- 图标10 分
- 文档15 分
- 许可证10 分
其他 Rmcp 集成
RMCP
dev.rmcp官方发布者
一个远程 MCP 目录服务器,让助手搜索已收录的 MCP 服务器与工具、查看其记录和端点,并比较候选方案。
YTDL RMCP
ai.dinglebear
本地运行的 Rust MCP 服务器,让助手搜索 YouTube、用 yt-dlp 下载音视频、写入标签并传输媒体,以及同步 Plex 播放列表。
Unraid RMCP
ai.dinglebear
本地 MCP 服务器兼 CLI,让助手通过 Unraid 的 GraphQL API 处理 NAS、Docker、虚拟机和存储相关任务。
Unraid RMCP
ai.dinglebear
本地运行的 Rust MCP 服务器兼命令行工具,让助手通过 Unraid 的 GraphQL API 查询和管理 NAS,涵盖阵列、Docker、虚拟机和存储。
UniFi RMCP
ai.dinglebear
本地 MCP 服务器兼 CLI,让助手读取 UniFi Network 控制器的客户端、设备、WLAN 和健康状态,并在具备管理员权限时执行变更操作。