The "multi-agent" topic on GitHub groups 113 open-source projects in the DeepSeek Harness (DSH) ecosystem, led by CowAgent with 46.9k GitHub stars. CowAgent — Open-source super AI assistant & Agent Harness. Every project here is indexed by DSH Universe with live GitHub data — stars, activity and install status — so you can compare and install directly.
The most complete, continuously updated, hottest collection of 100+ persona distillation skills|multi-agent system|celebrities/directors/Tianya legends/ancient texts/anime/workplace/relationships, all categories
📚 A zero-dependency, git-backed micro-lesson library for AI Agents to asynchronously share and search verified debugging experience. Python stdlib only. | https://misakanet.org
OPC — One Person Company. A full team in a single Claude Code skill. Adaptive agent orchestrator with 21 built-in roles, 6 flow templates, and adversarial quality control.
Open-source AI skills for SEO, AI search visibility, conversion copy, marketing strategy, and business operations. Reusable workflows for Claude Code and Codex, plus code review, app delivery, and creator tools.
Persistent multi-model workflow teams for DeepSeek Harness — dynamic lead planning, bounded DAGs, per-agent model/tools, Run Center and Token insights.
Bring Claude Code's UltraCode mode to DSH, upgrading DSH's one-shot multi-Agent scheduling into a generatable, savable, governable, observable, and recoverable Workflow layer
A visual multi-Agent workflow orchestration plugin built for the DeepSeek Harness Web GUI. Public beta is live! The official release will follow the DSH official release! Compatibility update: the plugin is now aligned with the latest release candidate dsh v0.1.2-rc.1, and does not support Alpha versions. Notice: since the plugin's underlying architecture relies on the official code's injection rules, there are no plans to support strongly breaking-change versions, because they are really hard to fix!
Interactive long-session background agents for DeepSeek Harness: start a durable continuable child agent, watch its progress in the Web UI sidebar, message it any time, and interrupt it - all through the official subagent seam.
Build your Agent kingdom in DSH——give Agents roles, permissions, territories, and governance order, and through fact-verification mechanisms and an authoritative ledger, make every execution, claim, and ruling verifiable and traceable.
Turn your devices into a fleet — dph plugin for decentralized multi-device collaboration: mDNS discovery, key pairing, SSH direct exec. Tools auto-register for dph agents.
dsh plugin: run work as a small crew of role agents (product manager, researcher, architect, engineer, QA, code reviewer, security reviewer, doc reviewer) that share work through files on disk
Role-based AI swarm orchestration for DeepSeek Harness: per-role model pinning with fallbacks, parallel task DAG with review loops, live Swarm dashboard tab.
Unified fleet-wide agent memory system for DeepSeek Harness — shared Obsidian Vault for dsh, Codex, Claude Code & Hermes with dependency-free Python FTS5 core.
Multi-agent orchestration for DeepSeek Harness: turn isolated sessions into a team that assigns work, reviews output, and runs DAG workflows — durable task ledger, real reviewers, crash-safe recovery.
Multi-agent collaboration suite for DeepSeek Harness: specialist roster with on-demand dispatch, roundtable, model comparison and a multimodal vision bridge — models via the official provider flow.
DSH multi-party group chat collaboration plugin (round table): pull AI Agents with different personas into the same discussion room to debate, cross-supplement, question and correct each other, then summarize a combined plan in one click; also supports pulling existing conversations in the workspace into the discussion to eliminate information silos. Supports three speaking modes — manual @ / all-hands discussion / relay chain — with round limits, global pause and other anti-infinite-loop controls. Official bundle plugin: dsh plugin --profile web add github:vuchisu069-source/dsh-round-table#main
One model writes the work, a different one reviews it, and neither decides whether it passed. Cross-provider AI review with verifiable evidence, for DeepSeek Harness/Cordis hosts. 中文文档见 README.zh.md。
Custom modes and custom prompts for DeepSeek Harness (dsh): edit a mode's system prompt in the settings page, pick its base mode, switch plugin rows one by one, and keep several assistants side by side — coding, chat or roleplay (RP) personas (multi-mode / multi-agent).
Watch Claude Code and Codex work inside dsh: each gets a real terminal pane in your session's workspace that you can type into, plus five tools so the dsh agent seats them, hands them work, and reads their screens.
A project-agnostic multi-agent collaboration protocol for the DeepSeek Harness: loop guards that stop runaway agent loops, a six-field handoff format, risk-tiered review routing, and a fixed verify→commit→push→review→close sequence. One skill instead of the same rules copy-pasted into every agent's instructions.
DeepSeek Harness multi-expert bundle: expert teams as user-authorable data directories, each auto-registered as a routable skill (engine and roster separated).
dsh-FEMO (Flow Emerges Opus) — dsh plugin multi-agent script engine: Femo sessions, chat window character bubbles, script control (full engine included, move the whole folder and it just works). Private until scrub review.
100% AI-native product delivery workflow plugin for DeepSeek Harness: full product-to-release pipeline (research → PRD → OpenSpec → parallel multi-agent → review loops → tests → release audit) with loop closure. Product delivery workflow plugin: full product-to-release pipeline, closed loop built in, 100% AI-native, start it before bed and collect the results when you wake up.
🌊 Abyss Agency · Abyss — paints DeepSeek Harness multi-agent runs as an agency you can see: delegation lineage, per-agent cost and failure attribution, context level, attendance timeline; past cases can be replayed and exported as a Markdown retrospective in one click. A DeepSeek Harness plugin that turns an agent fleet into an office you can watch.
DeepSeek Harness plugin: subagent_routed — run a one-shot subagent fully mounted on ANY agent preset from ANY session, with per-call model/provider override and a model-availability pre-check.
Captain Call - WeChat-style voice calls with your AgentTeams agents in DeepSeek Harness: GINKA desktop assistant, Kokoro open-source Chinese TTS, contact book with per-member voice picker. | Captain Call: WeChat-style voice call plugin with your Agent teammates
DeepSeek Harness multi-agent collaborative development plugin with seven built-in agents (Qibo, Lishou, Feng Hou, Kui, Li Mu, Gao Yao, Li Zhu), covering the full development loop of project initialization, code categorization, plan orchestration, batch execution, code review and test verification. Built on the @deepseek-ai/cordis plugin system, giving dsh "Feng Hou & Li Mu"-style multi-agent collaboration.
/code-review slash command for DeepSeek Harness — five parallel review lenses, per-finding confidence scoring, then a gh reply back on the pull request.
DeepSeek Harness named subagent roster plugin: settings-driven role roster, list_subagents / delegate tools and a settings page. Named subagent roster plugin for DeepSeek Harness.
Team rooms for DeepSeek Harness: persistent shared rooms across independent sessions - a message bus, a shared task board and a timeline. Extracted from dsh-background-agents, whose background-agent half is superseded by DSH's native continuable subagents.
Multi-agent collaboration shared workspace for DSH (DSH × Claude Code × Kimi Code × DevEco Code): outbox/status/tasks/knowledge/artifacts protocol as a cross-platform DSH plugin + CLI.
DSH plugin: AI multi-role collaborative development framework (files as communication) —— one-click init of collaboration convention files; any DSH session drives multi-role AI collaboration to complete software projects via ai_collab_init.
Knowledge Forge — a DeepSeek Harness native implementation of the Karpathy-style "LLM-compiled knowledge base": 5 pure-agent pipelines for compile/link/metabolize/Q&A archiving/health check, zero Python, zero vector DB, retrieval is thinking.
A DeepSeek Harness plugin that compiles a described world into declarative rules and lets agents, subjectivity, and emotion emerge — encode the story, not the emotion.
Peer-to-peer conversation linking for DeepSeek Harness (DSH): @-mention any session, wake it up, chat across conversations. Type @ in the composer, pick a session, send — the target agent is woken via agent.send. Persistent message board, native-feel inbox panel, 5 chat_link_* tools. dsh-plugin.
Mr. Meeseeks: heterogeneous free-model teams for DeepSeek Harness (dsh-agent-teams replica with Meeseeks art) + Budget-AGI brain-and-swarm preset + swarm scripts
DeepSeek Harness parallel multi-agent orchestration plugin - dispatch multiple sub-processes to run in parallel, main process can call in and revise anytime (AI-made, works decently)
Autonomous team collaboration plugin for DeepSeek Harness: captain splits tasks, specialized members with independent API/plugins run in isolated DSH subprocesses, JSONL mailbox IPC