AGPL-3.0 Android AI workspace for multi-provider chat, isolated assistants, local tools, Skills, MCP, translation, and backups; DSH-specific integration is not documented in the README.
DSH integration
Ecosystem-related
Safety audit
Unaudited
Last verified
2026-09-13
License
AGPL-3.0
01What can it help you accomplish?
Use one Android app for model services, assistants, tools, conversations, and local workspaces
A native Android AI workspace with provider connections, assistant configuration, organized conversations, and local files
Android users who want a local-first client for multiple model providers and assistant workflows
Connect different providers and give assistants isolated working contexts
OpenAI-compatible, Gemini, Claude, and Codex access plus per-assistant prompts, memory, tools, MCP servers, Skills, and history
Developers and power users comparing a multi-provider Android client for DSH-adjacent workflows
02How to install into DeepSeek Harness
Prerequisites
- Official releases target arm64-v8a devices running Android 8.0 or newer.
- A configured provider service or an eligible ChatGPT subscription is needed for model access.
Installation steps
- 01
Install the latest Miffan APK from GitHub Releases.
- 02
Open Settings → Providers and configure a service, or sign in to OpenAI Codex with a supported subscription.
- 03
Add or discover a model, then select it globally or for a specific assistant.
- 04
Enable search, MCP, Skills, local tools, or a workspace only when you need them.
- 05
For a source build, clone the repository and run ./gradlew assembleDebug.
Verify the integration
- Run ./gradlew test for the documented test task.
- Run ./gradlew lint for the documented lint task.
03DSH integration and capability boundaries
README-documented Android AI client; no direct DeepSeek Harness or dsh CLI integration is described.
Multi-provider model access
Provider credentials, endpoints, models, or an eligible ChatGPT subscription→Chat, reasoning, tool calls, image generation, and multimodal input according to the selected model
Requests depend on the provider service and account configured by the user.Assistant workspaces
Assistant prompts, model parameters, memory, tools, MCP servers, Skills, and conversation history→Independent assistant contexts with configurable prompts, tools, models, memory, and visual identity
Selected-text translation
Selected text from another Android app→Translation in a compact floating window without leaving the current page
Tools, Skills, and local workspaces
Enabled tools, Skills, MCP servers, or workspace configuration→Web search, device tools, isolated Linux workspaces, terminals, files, and AI file/shell tools
Capabilities can access external services or device data within their granted scope.Local data and portability
Conversations, settings, backups, and compatible provider data→Local storage, selective backup and restore, WebDAV/S3-compatible storage, and imports from other clients
Existing RikkaHub data is not shared automatically.
04Who is it for? When not to use it?
Good for
- Android users who want a local-first client for multiple model providers and assistant workflows
- Developers and power users comparing a multi-provider Android client for DSH-adjacent workflows
Not for
- Miffan does not bundle a model or replace a provider account; availability and charges depend on the services you configure.
- Prompts, attachments, and tool data are sent to the model, search, speech, MCP, sync, or other endpoints you choose.
- The documented official release lane is signed APKs for arm64-v8a devices running Android 8.0 or newer.
05Compatibility, maintenance and safety notes
- Miffan does not bundle a model or replace a provider account; availability and charges depend on the services you configure.
- Prompts, attachments, and tool data are sent to the model, search, speech, MCP, sync, or other endpoints you choose.
- The documented official release lane is signed APKs for arm64-v8a devices running Android 8.0 or newer.
AGPL-3.0 Android AI client documented in the repository README
06Frequently asked questions
Does Miffan document a native DeepSeek Harness or dsh integration?
No direct dsh CLI, plugin manifest, or DeepSeek Harness setup command is documented. The README describes Miffan as a native Android AI client and this record therefore classifies it as ecosystem-related rather than a native DSH plugin.
Can I install Miffan through a DSH plugin command?
No DSH installation flow is provided in the README. The documented distribution path is a signed APK from GitHub Releases; the source path uses `git clone`, `cd Miffan`, and `./gradlew assembleDebug`.
What DSH-adjacent model access does Miffan document?
The README documents OpenAI-compatible, Gemini, and Claude services, plus browser-based OpenAI Codex sign-in using access included with eligible ChatGPT subscriptions. It does not claim a DSH runtime connection.
How should I verify a source build before using it with my workflow?
The README lists `./gradlew test` and `./gradlew lint` as useful verification commands. It does not provide a DSH-specific health check or rollback command, so those fields remain undocumented here.
07Related DSH workflows
rea
by morluto
Reverse engineer anything with agents, from app behavior down to native binaries.
ruflo
by ruvnet
🌊 The original agent harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, federation, vector RAG integration, and native Claude Code / Codex / Hermes and many more Integrated
reactive-resume
by reactive-resume
A one-of-a-kind resume builder that keeps your privacy in mind. Completely secure, customizable, portable, open-source and free forever. Try it out today!
next-ai-draw-io
by dayuanjiang
Draw and edit draw.io diagrams by chatting with an AI. Web app, desktop app, and an MCP server for Claude Code, Cursor, Codex and VS Code.
08Data and sources
This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-09-13. Found an error? Submit a correction.
Best DeepSeek Harness Plugins
Twelve plugins worth installing first — picked from the whole catalog, across every category.