Apache-2.0 的 AI Agent 统一虚拟文件系统，把约 50 个服务挂载成一棵 bash 友好的目录树，可通过原生适配器、插件、MCP 或 FUSE 接入 DeepSeek Harness。
DSH 适配
兼容
作者声明
安全审计
未审计
最后核验
2026-08-21
许可证
Apache-2.0
01它能帮你完成什么？
Give DeepSeek Harness bash-level access to S3, Google Drive, Slack, Gmail, Redis and ~50 more services as one virtual filesystem
A single mounted workspace where read, grep and pipe commands sweep every backend out of the box, with zero new vocabulary
Developers running DeepSeek Harness (dsh) who want their agent to query real cloud and SaaS data without wiring up separate SDKs or MCP servers
Compose cross-service pipelines and keep agent runs portable on DeepSeek Harness
bash pipelines that span backends (e.g. run a script stored in Slack and file the report into Redis), plus clone / snapshot / version of portable workspaces that move between machines
Teams building agent workflows on DeepSeek Harness that need reproducible, portable workspaces across services
02如何接入 DeepSeek Harness？
前置条件
- Python ≥ 3.11 for the `mirage-ai` package and the `mirage` CLI
- Node.js ≥ 20 for the TypeScript SDK
- macOS or Linux (FUSE-based mounts require platform support)
安装步骤
- 01
Python: run `uv add mirage-ai` — installs the `mirage` library and the `mirage` CLI binary
- 02
TypeScript: run `npm install @struktoai/mirage-node` (Node.js servers and CLIs), `npm install @struktoai/mirage-browser` (browser / edge runtimes), and optionally `npm install @struktoai/mirage-agents` (OpenAI / Vercel AI / LangChain / Mastra adapters)
$ npm install @struktoai/mirage-node
- 03
CLI: run `curl -fsSL https://strukto.ai/mirage/install.sh | sh`, or `npm install -g @struktoai/mirage-cli`, or `uvx mirage-ai`, or `npx @struktoai/mirage-cli`
$ curl -fsSL https://strukto.ai/mirage/install.sh | sh
验证接入成功
作者未说明
03DSH 适配与能力边界
Used inside DeepSeek Harness as a coding-agent tool layer via native adapter, installable plugin, MCP, or FUSE mount
Unified virtual filesystem with ~50 built-in backends
S3 / R2 / GCS / Supabase, Gmail / GDrive / GDocs / GSheets / GSlides, GitHub / Linear / Notion / Trello, Slack / Discord / Email, MongoDB / Postgres / LanceDB / Qdrant, Redis, SSH, RAM, Disk→one filesystem tree under a single root, readable with standard bash semantics (read, grep, pipe)
reads and writes are executed against the mounted remote services using your own credentials/tokensPortable workspaces (clone / snapshot / version)
a configured Mirage workspace→snapshot archives (e.g. demo.tar) that can be loaded on another machine to restore the workspace
snapshot files are written to diskTwo-layer cache for remote backends
repeated reads against remote backends→index cache (listings/metadata, default TTL 10 minutes) and file cache (object bytes, default 512 MB) served from local state
cache state is held in-process RAM by default; an optional Redis store shares cache state across workers, processes, and machinesEmbeddable Python and TypeScript SDKs
a Python ≥ 3.11 or Node.js ≥ 20 application→workspaces running in-process inside FastAPI, Express, browser apps, or any async runtime — no separate process required
04适合谁？何时不该用？
适合
- Developers running DeepSeek Harness (dsh) who want their agent to query real cloud and SaaS data without wiring up separate SDKs or MCP servers
- Teams building agent workflows on DeepSeek Harness that need reproducible, portable workspaces across services
不适合
- Mirage targets macOS or Linux; FUSE-based mounts additionally require platform support, so FUSE mounting is not guaranteed on every machine.
- Mounted remote backends (Slack, S3, Gmail, Redis, ...) are reached over the network with your own credentials — e.g. a Slack bot token is passed in the resource config — so writes can mutate the underlying services.
05兼容性、维护与安全提示
- Mirage targets macOS or Linux; FUSE-based mounts additionally require platform support, so FUSE mounting is not guaranteed on every machine.
- The Python SDK and CLI require Python ≥ 3.11, and the TypeScript SDK requires Node.js ≥ 20 — older runtimes cannot run Mirage.
- Mounted remote backends (Slack, S3, Gmail, Redis, ...) are reached over the network with your own credentials — e.g. a Slack bot token is passed in the resource config — so writes can mutate the underlying services.
Apache-2.0 · actively maintained (latest release v0.0.5, 2026-08-15)
06常见问题
Mirage 如何接入 DeepSeek Harness？
DeepSeek Harness 被列在 Mirage 官方的编码代理集成表中，可通过原生适配器、可安装插件、MCP 或 FUSE 挂载接入；Mirage 文档中还有专门的 DeepSeek Harness 快速上手页（docs.mirage.strukto.ai/typescript/agents/dsh）。
是原生集成还是走 MCP？
两条路都有：编码代理可以通过原生适配器和可安装插件接入，也可以走 MCP 或 FUSE，按你在 DeepSeek Harness 中的工作流选择即可。
安装前需要准备什么？
`mirage-ai` 包和 `mirage` CLI 需要 Python ≥ 3.11，TypeScript SDK 需要 Node.js ≥ 20；如果要使用基于 FUSE 的挂载，还需要 macOS 或 Linux 且平台支持 FUSE。
可以挂载哪些服务？
内置约 50 个后端：RAM、Disk、Redis、S3 / R2 / OCI / Supabase / GCS、Gmail / GDrive / GDocs / GSheets / GSlides、GitHub / Linear / Notion / Trello、Slack / Discord / Email、MongoDB / GridFS / Postgres / LanceDB / Qdrant、SSH 等，全部并排挂载在同一个根目录下。
读写操作的数据去向是什么？
命令直接作用于挂载的后端：远程服务通过网络访问，使用你自己的凭据（例如 Slack bot token）；双层缓存（索引缓存 + 文件缓存，默认在进程内 RAM，可选 Redis）会让重复读取走本地而不重复访问网络。
07相关的 DSH 工作流
weknora
作者 tencent
开源的 LLM 知识库平台：把原始文档转化为可检索的 RAG 知识库、自主推理智能体和自动维护的 Wiki 系统，支持多租户部署。
honcho
作者 plastic-labs
用于构建有状态 AI 智能体的记忆库，支持跨会话连续学习与上下文工程。
reme
作者 agentscope-ai
ReMe：面向 AI Agent 的记忆管理套件——帮 Agent「记住我、优化我」，支持 RAG 与长期记忆。
memsearch
作者 zilliztech
为 Claude Code、Codex 等 AI Agent 提供持久化、统一的多层记忆：基于 Markdown 与 Milvus 向量库，支持语义检索与混合搜索。
08数据与来源
**Agent integrations:** OpenAI Agents SDK, Vercel AI SDK, LangChain, Pydantic AI, CAMEL, and OpenHands via the SDKs; cod…
| Coding agents | [Claude Code](https://docs.mirage.strukto.ai/python/agents/claude-code), [Codex](https://docs.mirage.s…
页面基于项目公开文档、仓库元数据和 DSH Plugins 的结构化解析生成；最后核验于 2026-08-21。发现错误？提交更正。
最佳 DeepSeek Harness 插件
从全目录挑出的 12 个值得优先安装的插件，覆盖各个分类。