DeepSeek Harness Plugin Hub

Publish and manage complete Harness Profiles. Discover Plugins for your next setup.

Explore

PluginsPresetsDocsNews

Community

Publish a pluginContactReport an issue

Resources

Plugin Hub on GitHubDeepSeek HarnessSystem statusPrivacy notice
© 2026 DeepSeek Harness Plugin HubPowered byPaxTech

Independent and unofficial. Not affiliated with, authorized by, or endorsed by DeepSeek.

DeepSeek Harness Plugin Hub
ProfilesPluginsCategoriesNewsDocsSign inManage Profiles
ProfilesPluginsCategoriesNewsDocsSign in

COMMUNITY CHALLENGES

Find your next challenge.

Real problems from the community. Find one that sparks an idea, explore the discussion, and build something people need.

Showing 61–80 of 286 community requestsUpdated hourly from community discussions
All updatesDSH ReleasesFeature Requests
All3 days7 days14 days
TopNew
OpenShippedSolved by pluginWon't fixAll statuses
#Problems worth solvingCommunity demand
61
HARNESS
2

Add image-input support for multimodal models

Support image inputs when connecting DeepSeek Harness to vision-capable models, so image requests are not incorrectly rejected as unsupported.

Explore challenge ↗

The problem

Support image inputs when connecting DeepSeek Harness to vision-capable models, so image requests are not incorrectly rejected as unsupported.

Possibly addressed by:Vision Subagent · Vision Recognizer · Multimodal Bridge

Start with these discussions

  • I run DeepSeek-V4.1-Flash-1017 across locally four DGX Sparks , on a 200G switch, behind my own gateway, with pi and Deex · ▲1 · 1 comments
  • 我用Deepseek harnessx · ▲0 · 1 comments
Last active Oct 31, 2026
2 authors2 posts · 1 reactions2 comments
62
HARNESS
6

Run Persistent SSH Agent Workspaces

Run the complete agent runtime, sessions, tools, plugins, and background work on an SSH host so state survives client disconnects and can be securely resumed with durable event and approval recovery.

Explore challenge ↗

The problem

Run the complete agent runtime, sessions, tools, plugins, and background work on an SSH host so state survives client disconnects and can be securely resumed with durable event and approval recovery.

Possibly addressed by:Ssh Workspaces · Remote Workspaces · Remote Workspace

Start with these discussions

  • 支持SSH远程服务器工作区，Agent可在远端服务器环境执行开发任务（Full SSH Remote Workspace with Persistent Remote Agent Runtime）github · ▲3 · 3 comments
  • 什么时候把远程工作区的功能端上来啊？github · ▲2 · 0 comments
  • Remote workspaces in the Desktop app — and the remote side must keep running on its owngithub · ▲1 · 0 comments
Last active Sep 30, 2026
63
HARNESS
2

Persist Sidebar browser sessions across restarts

Make each workspace’s Sidebar browser use a stable persistent partition so cookies, local storage, and site logins survive desktop app restarts. This capability must be implemented in the harness because plugins cannot control the guest partition.

Explore challenge ↗

The problem

Make each workspace’s Sidebar browser use a stable persistent partition so cookies, local storage, and site logins survive desktop app restarts. This capability must be implemented in the harness because plugins cannot control the guest partition.

Possibly addressed by:Sidebar Manager · Session Folders · Session Isolation

Start with these discussions

  • Sidebar browser guests lose site logins on restart: give the workspace partition a persist: identitygithub · ▲1 · 1 comments
  • [Proposal] 侧边栏 Browser 的存储分区应可持久化：第三方站点登录态每次 App 重启后丢失github · ▲1 · 0 comments
Last active Oct 27, 2026
2
64
HARNESS
4

Configure per-conversation session-affinity headers

Attach stable per-conversation session identifiers, such as UUID-based OpenCode headers, to outbound inference requests. Allow provider- and model-level control over header names, formats, and omission, together with related prompt-cache settings to improve routing and cache affinity.

Explore challenge ↗

The problem

Attach stable per-conversation session identifiers, such as UUID-based OpenCode headers, to outbound inference requests. Allow provider- and model-level control over header names, formats, and omission, together with related prompt-cache settings to improve routing and cache affinity.

Possibly addressed by:Opencode Session Header · X Opencode Session · Session Header

Start with these discussions

  • Please send x-opencode-session header on API requests (OpenCode Go requires it from 09/05)github · ▲187 · 47 comments
  • Poor prompt-cache hit rate on z.ai GLM Coding Plan: prefix is stable, hits still collapse (evidence + proposal to expose pi-ai session-affinity switches for the zai routes)github · ▲1 · 2 comments
65
HARNESS
4hackernews1

Load tool definitions on demand

Defer and selectively inject tool descriptions and schemas only when the model needs them or the available tool set changes, reducing prompt, latency, and caching overhead while retaining tool access.

Explore challenge ↗

The problem

Defer and selectively inject tool descriptions and schemas only when the model needs them or the available tool set changes, reducing prompt, latency, and caching overhead while retaining tool access.

Possibly addressed by:Lazy Tools · Tool Visibility · Tool Visibility

Start with these discussions

  • 模型对话是否每一次都有必要带上工具定义？github · ▲2 · 3 comments
  • Performance: zero prompt-cache hits + large tool-schema overhead slow down long tasksgithub · ▲2 · 1 comments
  • [Feature Request] 工具按需加载（Tool on-demand loading）：降低多插件 profile 每轮 tool schema 注入成本
66
PLUGIN
6

Configure SSRF-Safe Fake-IP Resolution Allowlisting

Allow web-fetch plugins to configure trusted resolvers, domains, or CIDR ranges for transparent-proxy fake-IP DNS while retaining SSRF protection. Support relevant IPv4 and IPv6 pools, keep private and link-local networks blocked by default, and provide actionable errors.

Explore challenge ↗

The problem

Allow web-fetch plugins to configure trusted resolvers, domains, or CIDR ranges for transparent-proxy fake-IP DNS while retaining SSRF protection. Support relevant IPv4 and IPv6 pools, keep private and link-local networks blocked by default, and provide actionable errors.

Possibly addressed by:Web Fetch Enhanced · Fakeip Compat · Fetch Third Party

Start with these discussions

  • web-fetch-http: add a config toggle to allow non-public DNS answers (fake-ip proxy compatibility)github · ▲15 · 6 comments
  • web-fetch-http SSRF guard rejects every fetch under transparent-proxy fake-ip DNS (198.18.0.0/15) - no resolver/allowlist escape hatchgithub · ▲13 · 12 comments
  • 建议为 web_fetch 增加可配置的非公网 CIDR 白名单，以支持 Fake-IP 与特殊路由环境
67
HARNESS
4

Build robust provider-aware context management

Provide proactive, configurable compaction with model-visible pressure signals, bounded tool-output spilling, and repeat-call limits. Base it on actual provider limits and ensure deterministic progress with model-free fallbacks when summarization or oversized requests fail.

Explore challenge ↗

The problem

Provide proactive, configurable compaction with model-visible pressure signals, bounded tool-output spilling, and repeat-call limits. Base it on actual provider limits and ensure deterministic progress with model-free fallbacks when summarization or oversized requests fail.

Possibly addressed by:Smart Compaction · Handoff Compaction · Improved Compact

Start with these discussions

  • token 消耗过大的根因分析及改进建议github · ▲2 · 6 comments
  • [Proposal] Provider-neutral prompt token pressure + local overflow -> compact-retrygithub · ▲1 · 6 comments
  • Compaction cannot rescue an oversized session — the summarization request itself exceeds the provider request-size limit (HTTP 413)
68
HARNESS
2

Preserve unsent message drafts when switching conversations

Preserve partially written, unsent messages in new conversations when users switch to another conversation and return, so their drafting workflow is not interrupted.

Explore challenge ↗

The problem

Preserve partially written, unsent messages in new conversations when users switch to another conversation and return, so their drafting workflow is not interrupted.

Possibly addressed by:Pending Session · Session Drafts · Draft Sessions

Start with these discussions

  • [feature]希望新对话可以保留未发送出去的草稿github · ▲1 · 4 comments
  • 新对话草稿无法跨「新对话」保留github · ▲1 · 0 comments
Last active Oct 21, 2026
2 authors2 posts · 2 reactions4 comments
69
HARNESS
11

Add MCP support to DeepSeek Harness

Add support for the Model Context Protocol (MCP) to DeepSeek Harness so it can integrate with MCP-compatible tools and services.

Explore challenge ↗

The problem

Add support for the Model Context Protocol (MCP) to DeepSeek Harness so it can integrate with MCP-compatible tools and services.

Possibly addressed by:Harness Mcp Server · Mcp Manager · Mcp Link

Start with these discussions

  • deepseek-harness mcpgithub · ▲1 · 3 comments
  • @ask_dba DeepSeek harness doesn't supports MCP 😞😢x · ▲0 · 1 comments
Last active Oct 25, 2026
2
70
HARNESS
2

Surface startup failures with fallback logging

Show a user-facing error and write a startup log when the user-data directory, single-instance lock, or profile setup fails, including the path and OS error code. Ensure failures never exit silently by logging to a writable fallback location such as the system temporary directory, and optionally warn about restrictive ACLs or integrity levels.

Explore challenge ↗

The problem

Show a user-facing error and write a startup log when the user-data directory, single-instance lock, or profile setup fails, including the path and OS error code. Ensure failures never exit silently by logging to a writable fallback location such as the system temporary directory, and optionally warn about restrictive ACLs or integrity levels.

Possibly addressed by:Startup Check · Startup Check

Start with these discussions

  • Desktop app exits silently with exit code 3 when it cannot create its user-data files (no dialog, no log)github · ▲1 · 1 comments
  • Desktop app exits silently with code 3 when it cannot create its user-data files (no dialog, no log)github · ▲1 · 0 comments
Last active Nov 3, 2026
71
HARNESS
2

Refresh MCP tools after reconnect

Allow long-lived sessions to re-fetch and register MCP tools after a connection failure or successful reconnect, so recovered servers can be used without restarting the session.

Explore challenge ↗

The problem

Allow long-lived sessions to re-fetch and register MCP tools after a connection failure or successful reconnect, so recovered servers can be used without restarting the session.

Possibly addressed by:Live Mcp · Mcp Skill Control · Mcp Skill Control

Start with these discussions

  • mcp-client: stateful MCP server restart leaves every tool call failing until DSH restartsgithub · ▲2 · 5 comments
  • Headless session env: are DSH_-prefixed variables intentionally dropped before reaching the session?github · ▲1 · 1 comments
Last active Oct 15, 2026
2 authors2 posts · 3 reactions6 comments
72
HARNESS
2

Add native download actions to deliverable cards

Add an authenticated, size-capped file download action to deliverable cards and optionally changed-file rows, enabling direct downloads on headless or remote hosts. Also consider exposing card actions as a plugin extension point.

Explore challenge ↗

The problem

Add an authenticated, size-capped file download action to deliverable cards and optionally changed-file rows, enabling direct downloads on headless or remote hosts. Also consider exposing card actions as a plugin extension point.

Possibly addressed by:Download Button · Delivery Native Open · Save Button

Start with these discussions

  • Feature: native Download action on present deliverable cards (critical for headless/WSL hosts)github · ▲1 · 1 comments
  • Feature idea: direct "Download" action on deliverable cardsgithub · ▲1 · 0 comments
Last active Nov 2, 2026
1 authors2 posts · 2 reactions
73
PLUGIN
41

Configure Per-Request Timeouts

Support configurable longer timeouts for large-payload RPCs and model requests, including slow local model loading, generation, retries, and large outputs, while retaining shorter defaults for ordinary calls.

Explore challenge ↗

The problem

Support configurable longer timeouts for large-payload RPCs and model requests, including slow local model loading, generation, retries, and large outputs, while retaining shorter defaults for ordinary calls.

Start with these discussions

  • llm-pi-ai Timeout increase for local LLMsgithub · ▲18 · 21 comments
  • 使用本地部署的模型时，如果重启过电脑，Vram内容被清空，则发起对话时模型要思考很久时间，如果超过五分钟就报错重试了，导致无法继续对话github · ▲1 · 4 comments
  • dsh-client-connection：history 等大载荷 RPC 受全局 30s 超时限制，Web 端报「历史加载失败：signal timed out (internal)」—— timeoutPolicy 钩子已存在但无人接线github · ▲1 · 1 comments
74
HARNESS
4hackernews1

Support Workspace Relocation and Rebinding

Provide a supported workflow to rebind workspaces after folder moves or renames while preserving sessions and updating related paths, configuration, history, metadata, and caches. Safely isolate corrupt or mismatched sessions so relocation does not prevent startup.

Explore challenge ↗

The problem

Provide a supported workflow to rebind workspaces after folder moves or renames while preserving sessions and updating related paths, configuration, history, metadata, and caches. Safely isolate corrupt or mismatched sessions so relocation does not prevent startup.

Possibly addressed by:Session Workspace · Workspace Mover · Session Mover

Start with these discussions

  • Renaming a workspace folder makes its stored sessions fail the identity check, with no supported repairgithub · ▲2 · 2 comments
  • [Bug] 项目目录迁移后 dsh 启动崩溃（corrupt session log），希望支持会话重新归属
75
HARNESS
4hackernews2

Let users choose models and reasoning per task

Provide task-aware model selection with a provider-organized picker and an independent reasoning-effort control. Make the selection available in Composer and workflow dialogs, with persistence and keyboard-accessible controls.

Explore challenge ↗

The problem

Provide task-aware model selection with a provider-organized picker and an independent reasoning-effort control. Make the selection available in Composer and workflow dialogs, with persistence and keyboard-accessible controls.

Possibly addressed by:Model Selector · Effort · Model Select Plus

Start with these discussions

  • 切换模型的时候能不能别默认恢复推理等级？github · ▲2 · 2 comments
  • [Feature Proposal] Direct reasoning-effort slider beside the composer model selectorgithub · ▲2 · 1 comments
  • [Feature] Provider-first two-step model picker in the composer (plus a direct reasoning-effort slider)
76
PLUGIN
2hackernews42

Provide Remote-Control Clients for DSH

Provide polished desktop and mobile clients for connecting to remote DSH servers, monitoring and steering agents, reviewing work, taking over when needed, and returning conveniently to terminal workflows.

Explore challenge ↗

The problem

Provide polished desktop and mobile clients for connecting to remote DSH servers, monitoring and steering agents, reviewing work, taking over when needed, and returning conveniently to terminal workflows.

Possibly addressed by:Mobile Remote · Phone Remote · Remote X

Start with these discussions

  • 桌面端和手机端远程连接服务器上的dsh-web ?github · ▲1 · 2 comments
77
HARNESS
2

Add Linux x64 desktop packages

Provide unsigned Linux x64 desktop packages in .deb and AppImage formats through the existing packaging pipeline, since downloads currently support only macOS and Windows. Include the required Linux runtime and packaging adjustments for reliable operation.

Explore challenge ↗

The problem

Provide unsigned Linux x64 desktop packages in .deb and AppImage formats through the existing packaging pipeline, since downloads currently support only macOS and Windows. Include the required Linux runtime and packaging adjustments for reliable operation.

Possibly addressed by:Desktop Packager · Linux Desktop · Desktop Kit

Start with these discussions

  • Linux (x64) desktop packaging — working build and a minimal patch seriesgithub · ▲1 · 0 comments
  • Proposal: Linux x64 desktop packages (.deb + AppImage) — implementation available (fork branch)github · ▲1 · 0 comments
Last active Oct 26, 2026
2 authors2 posts · 2 reactions
78
PLUGIN
5

Decouple web search providers from the conversation model

Allow web search to use independently configured providers, credentials, endpoints, and optional auxiliary models, including a zero-LLM mode and reuse of the selected model when needed. Clearly expose provider failures and any search-related model charges, with appropriate safeguards and cost tracking.

Explore challenge ↗

The problem

Allow web search to use independently configured providers, credentials, endpoints, and optional auxiliary models, including a zero-LLM mode and reuse of the selected model when needed. Clearly expose provider failures and any search-related model charges, with appropriate safeguards and cost tracking.

Possibly addressed by:Web Search Diy · Web Search · Websearch

Start with these discussions

  • web_search 在会话模型切到第三方 provider 后仍走 DEEPSEEK_API_KEY 独立计费——披露藏在参考页一行里，且从不说破计费后果github · ▲2 · 4 comments
  • [DSH Desktop] web_search 联网搜索工具静默使用官方 DeepSeek API 并计费，且界面无任何提示github · ▲2 · 2 comments
  • 标题： [Feature Request] web_search 工具应支持不经大模型的纯搜索模式（当前每次搜索都调用付费 DeepSeek API）
79
HARNESS
4

Implement scoped, persistent agent memory

Provide a backend-agnostic, first-party memory plugin or service with session, agent, tenant, topic, and repository scoping; automatic capture and recall; safe lifecycle handling; provenance, deduplication, bounded injection, and tools for managing facts and lessons. Support importing existing memories and preserving history safely across migrations.

Explore challenge ↗

The problem

Provide a backend-agnostic, first-party memory plugin or service with session, agent, tenant, topic, and repository scoping; automatic capture and recall; safe lifecycle handling; provenance, deduplication, bounded injection, and tools for managing facts and lessons. Support importing existing memories and preserving history safely across migrations.

Possibly addressed by:Agent Memory · Plugin Memory · Hybrid Memory

Start with these discussions

  • dsh-plugin: Native OpenViking memory for DeepSeek Harness (automatic session capture + memory tools + profile injection)github · ▲2 · 5 comments
  • 给 #14 补点料：两个现成记忆实现里踩过的坑github · ▲1 · 6 comments
80
HARNESS
2

Add built-in theme palettes and a settings picker

Add first-class Aurora, Ember, Mint, and Default theme palettes with light/dark variants, a profile-persisted picker under Settings > General, and live switching without restart. This would make customization discoverable and consistent across the web GUI and desktop shell.

Explore challenge ↗

The problem

Add first-class Aurora, Ember, Mint, and Default theme palettes with light/dark variants, a profile-persisted picker under Settings > General, and live switching without restart. This would make customization discoverable and consistent across the web GUI and desktop shell.

Possibly addressed by:Theme Palette · Material Theme · Prism Themes

Start with these discussions

  • Built-in colour themes (Aurora, Ember, Mint) with a palette picker under Settings > Generalgithub · ▲1 · 0 comments
  • Built-in colour themes (Aurora, Ember, Mint) with a palette picker under Settings > Generalgithub · ▲1 · 0 comments
Last active Nov 3, 2026
← PreviousPage 4 of 15Next →
6 authors6 posts · 8 reactions4 comments
authors
2 posts · 2 reactions
1 comments
  • bug(llm-pi-ai): openai-responses 路由无法关闭 session_id 请求头github · ▲1 · 2 comments
    • Last active Sep 10, 2026
    4 authors4 posts · 190 reactions52 comments
    github · ▲1 · 1 comments
    Last active Oct 2, 2026
    5 authors5 posts · 6 reactions5 comments
    github · ▲10 · 3 comments
    Last active Sep 15, 2026
    6 authors6 posts · 44 reactions22 comments
    github · ▲1 · 3 comments
    Last active Oct 1, 2026
    4 authors4 posts · 5 reactions15 comments
    authors
    2 posts · 1 reactions
    4 comments
    1 authors2 posts · 2 reactions1 comments
    1 comments
    Last active Sep 19, 2026
    4 authors5 posts · 21 reactions27 comments
    github · ▲2 · 2 comments
  • 工作区文件夹绝对位置改变后工作区重绑定功能需求github · ▲2 · 1 comments
    • Last active Sep 27, 2026
    5 authors5 posts · 7 reactions5 comments
    github · ▲1 · 1 comments
    Last active Sep 26, 2026
    6 authors6 posts · 6 reactions4 comments
    win,安卓客户端连接 dsh web
    github · ▲1 · 0 comments
  • Okay so I'm spoiled by ChatGPT Codex Remote Control. What compares in the OSS world?x · ▲0 · 1 comments
    • Last active Oct 2, 2026
    6 authors8 posts · 2 reactions3 comments
    0 comments
    github · ▲2 · 1 comments
    Last active Sep 22, 2026
    5 authors5 posts · 9 reactions8 comments
  • 是否可以支持多租户记忆隔离，提供详细的api文档github · ▲1 · 3 comments
    • Last active Sep 25, 2026
    4 authors4 posts · 5 reactions15 comments
    1 authors
    2 posts · 2 reactions
    0 comments