lobehub vs mateclaw
LobeChat's 80k-star pivot: from chat UI to 'Chief Agent Operator' — hire, schedule and supervise a team of agents running 7×24, self-hosted via Docker or Vercel, plugin ecosystem carried over. — versus — Self-hosted Java agent platform: ReAct plus plan-and-execute 'digital employees' with an LLM Wiki knowledge layer, skills/MCP/ACP tools, approval-gated actions and eight IM channels in one JAR.
Both are self-hosted platforms for hiring and supervising a team of always-on agents; LobeHub is the TypeScript/Docker/Vercel path with a huge plugin ecosystem, MateClaw the JVM one with RBAC, approval gates and IM channels built in.
| lobehub | mateclaw | |
|---|---|---|
| Stars | 82k | 1.0k |
| Forks | 16k | 318 |
| Language | TypeScript | Java |
| License | NOASSERTION | Apache-2.0 |
| Last activity | today | today |
| Topics | agents, orchestration | agents, orchestration, memory |
| Curated connections | 5 | 5 |
lobehub — the curator's take
One of the biggest chat-UI codebases reinventing itself as an agent-team operator — hire, schedule, report, stay in charge without staying online. Read the star count correctly: 80k stars measure the LobeChat era, not the operator bet, which is weeks old and under active rebuild — evaluate those features on their own merits. What carries over is real: mature self-hosting (Docker/Vercel/Zeabur), a large plugin ecosystem, and an unusually strong design culture. GitHub reports no standard license (custom terms) — read them before commercial deployment; this is not plain MIT anymore. When NOT: if you need a proven production agent runtime today, the harness layer is further along than the operator layer.
mateclaw — the curator's take
Pick it when the buyer is an IT department rather than a hacker: multi-user workspaces, RBAC and approval gates on sensitive tool calls, audit trail, Spring Actuator health, per-channel error isolation, and provider failover across DashScope/OpenAI/Anthropic/Gemini/DeepSeek/Ollama when a vendor 401s mid-request. Underneath it's a real harness — StateGraph runtime, Team Runs with a durable runId, task DAGs, human approval — not a RAG call in a wrapper. The cost is weight and gravity: Java 21 + Spring Boot 3.5 + Vue, an enormous and fast-moving feature surface (workflows, triggers, Wiki transformations, Dream consolidation, skill mining), and a China-first channel set (DingTalk, Feishu, WeChat Work, QQ) whose value drops sharply outside that ecosystem. For a personal agent on a $5 VPS this is orders of magnitude too much machinery.