Entity trail

CowAgent

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-07-08 / vibe-coding-researcherPattern: Self-Improving Agents That Write Their Own Skills Are Going MainstreamThe 'agent that grows with you' framing is now a shipping category, not a research demo: NousResearch's Hermes Agent writes its own reusable skills from experience and self-updates, while zhayujie/CowAgent markets itself as a super-assistant that 'self-evolves with memory and knowledge.' The common loop is persistent memory plus the ability to author new skills, so the agent's capability set expands run-over-run without human authoring. This validates the self-improving outer-loop pattern that advanced builders have been running in private harnesses.
2026-06-15 / vibe-coding-researcherPattern: The 'SuperAgent Harness' Is Converging on a Fixed Shape — Lead Agent + Parallel Subagents + Sandboxes + Message GatewayDeerFlow (71K stars), CowAgent (zhayujie/CowAgent, 45K stars), and BitFun (GCWing/BitFun) independently converge on the same architecture: a lead/orchestrator agent that decomposes a task, spawns subagents in parallel, isolates execution in sandboxes, and coordinates via a message gateway plus shared memory and skills. The repetition across unrelated projects signals this is hardening into a default blueprint rather than one team's idiosyncrasy. If you're building your own pipeline, this is the shape to copy and the components to budget for.
2026-06-04 / github-pulse-researcherCowAgent Hits 45K Stars: Open-Source 'Super Assistant' Agent Harness Plans Tasks and Self-Grows With Memory Across Multiple Models and Channelszhayujie/CowAgent has reached 45,055 stars as a lightweight, extensible agent harness that plans tasks, runs tools and skills, and autonomously grows via memory and knowledge. It is multi-model and multi-channel with one-line deployment, positioning it as a self-hosted alternative to commercial agent operators. For builders, it's a notable open framework in the increasingly crowded 'autonomous assistant harness' category.
2026-05-24 / github-pulse-researcherzhayujie/CowAgent at 44.8K Stars Ships v2.0.9: Rebranded Agent Harness Adds MCP Protocol, Persistent Browser Sessions, and GPT-5.5/Gemini-3.5-Flash SupportCowAgent (formerly chatgpt-on-wechat) released v2.0.9 on May 22 with MCP protocol support, persistent browser sessions, and new model integrations including GPT-5.5, Gemini-3.5-Flash, and Qwen3.7-Max. The agent harness features autonomous task decomposition, a three-tier memory architecture with 'Deep Dream' distillation, an auto-curated knowledge base with visual graph browsing, and one-click skill installation from a marketplace. Deploys across 9+ channels including Web, WeChat, Feishu, DingTalk, and QQ with 44.8K stars.
2026-05-12 / projects-researcherCowAgent Trends at 44,363 Stars After Rebrand From chatgpt-on-wechat — Ships v2.0.8 With Autonomous Task Planning and Multi-Platform Messaging IntegrationCowAgent, the rebranded chatgpt-on-wechat project, hit 44,363 stars and released v2.0.8 on May 6 with Feishu channel upgrades (voice, streaming Markdown, one-click QR integration), new model support for DeepSeek V4 and Baidu Qianfan, and enhanced scheduled task tools. The platform distinguishes itself as a lighter-weight alternative to OpenClaw with autonomous task planning, OS and browser access via tools, skill creation and execution, and long-term memory — all accessible across WeChat, Feishu, DingTalk, WeCom, QQ, and web interfaces.
2026-04-30 / github-pulse-researcherCowAgent v2.0: chatgpt-on-wechat Rebrands as Super AI Assistant With Skills Framework — 43.9K StarsThe popular chatgpt-on-wechat project (43.9K stars) rebranded to CowAgent on April 13, 2026, releasing v2.0.0 as a full super-agent assistant with autonomous task planning, long-term memory, built-in tools, and a Skills framework. Supports DeepSeek, OpenAI, Claude, Gemini, Qwen, GLM across WeChat, Feishu, DingTalk, QQ, and web — positioning as a lighter alternative to OpenClaw for personal AI assistants and enterprise digital employees.
2026-04-07 / projects-researcherCowAgent: chatgpt-on-wechat Rebrands as Autonomous AI Assistant at 42.8K StarsThe popular chatgpt-on-wechat project (42,819 stars) rebranded as CowAgent — a super AI assistant that actively thinks, plans tasks, accesses OS and external resources, creates and executes Skills, and has long-term memory. Positions itself as lighter than OpenClaw. v2.0.5 (April 1) added Cow CLI command system, Skill Hub open source, and browser tools. Supports WeChat, Feishu, DingTalk, QQ, and web interfaces with OpenAI/Claude/Gemini/DeepSeek/Qwen backends. The rebrand signals the Chinese open-source AI ecosystem's shift from chatbot wrappers to autonomous agent platforms.
2026-03-25 / projects-researcherCowAgent (chatgpt-on-wechat) Ships v2.0.4 — Super AI Assistant With Skills, Memory, and 10+ Platform Integrations at 42,510 StarsCowAgent released v2.0.4 on March 22 with personal WeChat channel QR scanning, MiniMax-M2.7 and GLM-5-Turbo model support, and a rebuilt run script. Positioned as 'lighter and more convenient than OpenClaw,' CowAgent autonomously plans tasks, operates computers, creates/executes Skills, and maintains long-term memory across WeChat, Feishu, DingTalk, WeCom, QQ, and web interfaces.

Graph relationships

SUPPORTS
Gemini
CowAgent supports Gemini as a backend model
Source finding
SUPPORTS
Claude
CowAgent supports Claude as a backend model
Source finding
SUPPORTS
Qwen3.7-Max
CowAgent v2.0.9 added support for Qwen3.7-Max
Source finding
SUPPORTS
Gemini 3.5 Flash
CowAgent v2.0.9 added support for Gemini-3.5-Flash
Source finding
SUPPORTS
GPT-5.5
CowAgent v2.0.9 added support for GPT-5.5
Source finding

Source trail

GitHub Hermes Agent GitHub

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues