Agents publish interactive pages people comment on, approve and fill in; agents read the answers.

Generative UI for DeepSeek Harness — the agent writes TSX and the web UI compiles every frame as it streams, so a card renders while its JSX is still unclosed. Inline in the conversation, or in a side canvas.

Give an AI a research direction: it runs literature synthesis, gap validation, hypothesis formation and experiment design itself. 267 markdown skills form a research graph: 51 operations built from 216 atomic steps, any order, explicit backtracking; the plan is approved before it runs and interrupted runs resume. npx install, no runtime, no API key

[dsh]为纯文本模型设计更强大的视觉工具箱：一行安装使用、粘贴图片直接识别、多张图片问答、截图到前端UI 还原等｜DeepSeek Harness-native integration for agent-vision-toolkit: image Q&A, long-screenshot OCR, UI restoration, grounding, pixel diff, Artifacts, and Web UI.

DeepSeek Harness Desktop App: a local AI desktop workspace for DSH Sessions, projects, files, web research, plugins, and Office artifacts.

Reusable project context and specialist workflows for the AI tools you already use. AIWG places agents, skills, commands, and rules in provider-readable locations, with optional utilities for artifact memory, workflow orchestration, recovery, and discover

Evidence-backed UI delivery for coding agents working on existing Web products: declared criteria, criterion-bound artifacts, findings linked to declarations, and repair guidance. Evaluator review does not replace human semantic approval.

Reusable project context and specialist workflows for the AI tools you already use. AIWG places agents, skills, commands, and rules in provider-readable locations, with optional utilities for artifact memory, workflow orchestration, recovery, and discover

DSH Plugin Radar — open-source ecosystem radar for DeepSeek Harness plugins: continuous discovery (21k+ candidates), k8s runtime validation (13k+ tests), 15-min snapshots; the catalog is a generated artifact — 开源 DSH 插件生态雷达：持续发现 2.1 万+ 候选、k8s 运行级实测 1.3 万+、15 分钟快照；插件目录为自动生成的产物

Shared server-side core for MulmoClaude and MulmoTerminal — the always-shipped-together subsystems consolidated behind subpath exports so the two hosts can't drift. Server-only except the browser-safe ./artifacts, ./whisper/client, ./workspace-setup/slug,

A market-driven Product Manager copilot (Agent Skills + Claude Code plugin) based on the Pragmatic Framework — guides PMs from idea to launch and produces a client-ready artifact (Markdown or .docx) at every stage.

Exposes an assess_repo tool that scans a repository, local folder, or npm/PyPI package and returns supply-chain risk findings with file, line, and code evidence — install-time scripts, dangerous patterns, secrets, AI-agent manipulation, impersonation and published-artifact mismatches — while never executing the target's code. Findings carry severities from info to critical so results can gate installs and CI jobs.

Exposes three read-only tools that list local Agent and MCP runtime candidates, inspect running processes, and check a live runtime against an approved artifact expectation. Returns explicit verdicts such as matched, stale, leaked, conflict, unknown, or not running without exposing argv, environment values, or file contents.

Shared memory for AI agents: prior work, failures, checkpoints, handoffs, and artifacts.