DeepSeek Harness Vision Plugins

One-stop manager for skills, subagents, MCP services, and LSP servers in DeepSeek Harness.

A DeepSeek Harness provider plugin that transcribes attached images with an OpenAI-compatible vision model before sending text to DeepSeek.

Local document parsing and offline OCR tools for DeepSeek Harness.

Academic research agent skills bundle for literature search, PDF conversion to Markdown, and paper deep-reading.

Embedded multi-protocol AI routing gateway and model fallback manager for DeepSeek Harness Web.

Local-first literature RAG plugin for DeepSeek Harness with exact passage citations and Zotero integration.

An IDE-style Git panel and branch pill for the DeepSeek Harness Web GUI with diff viewers and AI commit suggestions.

Enables Codex-style clickable inline file paths and file manager shortcuts in DeepSeek Harness web replies.

An in-app browser pane and local file previewer for DeepSeek Harness with model-facing preview tools.

Persistent memory plugin for DeepSeek Harness backed by a local Obsidian Markdown vault.

Codex-style code review plugin for DeepSeek Harness using independent subagents to inspect Git diffs and commits.

End-to-end multi-agent team R&D pipeline plugin for DeepSeek Harness with kanban workbench and breakpoint resumption.

Real-time model token usage, cost tracking, and provider quota monitoring for DeepSeek Harness Web UI.

Merge multiple directories into a single DSH session with custom file tools and UI panels.

Noname extension development workshop with validation gates, source search, and P2P online play assistant.

LLM-assisted auto-approval pipeline and timeout fallback manager for DeepSeek Harness.

DeepSeek Harness settings extension for non-destructive per-model reasoning-effort levels, input modalities, and overrides.

Agent2Agent (A2A) v1.0 protocol dual-mode integration for DeepSeek Harness, supporting outbound skill-to-tool mapping and inbound JSON-RPC/SSE server endpoints.

Side-by-side multi-panel chat interface for DeepSeek Harness to view and interact with multiple agents simultaneously.

Adds configurable vision-model image-to-text conversion, image analysis tools, and a vision-model selector to DeepSeek Harness.

Import skills, MCP servers, instructions, and hooks from Claude Code, Codex, and OpenCode into DeepSeek Harness.

Routes auxiliary tasks like vision, context compaction, subagents, and image generation to dedicated models without altering the main chat model.

IDE-grade Git tool window for DeepSeek Harness with branch trees, commit graph, and chat insertion.

Local-first AI image design canvas workbench for DSH with image generation, smart editing, and Adobe round-trip bridging.

A pi-style terminal UI bundle for DeepSeek Harness with subagent steering, session history, and custom themes.

GitHub Copilot credential provider and OAuth device-flow sign-in integration for DeepSeek Harness.

Connect browser-logged-in web AI models (DeepSeek, GLM, Kimi, Qwen, Doubao) into DeepSeek Harness without API keys.

Dynamic wallpaper background plugin for DeepSeek Harness Web UI supporting Wallpaper Engine .mpkg, Steam Workshop, video, and web backgrounds.

Context quality and memory governance layer for AI agents with DeepSeek Harness integration and MCP support.

Live interactive DAG visualization panel for subagents, Agent Teams tasks, communication, and Workflows in DeepSeek Harness Web.

File upload and attachment manager plugin for the DeepSeek Harness web composer with model-aware prompt adaptation.

Community plugin marketplace UI for DeepSeek Harness with one-click installation and management.

DeepSeek Harness WebUI authentication bundle that gates web resources, plugin bundles, API RPC, and WebSocket connections behind login.

Zero-cost onboarding and model management bundle for DeepSeek Harness with free model setups, quota monitoring, and peak pricing reminders.

VS Code-style file explorer, source control, diff editor, and integrated terminal for the DSH Web GUI sidebar.

Template-driven AI code review skill for DeepSeek Harness with 5-step flow and dual HTML/text output.