lobehub vs qm
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 — Multiplayer agent harness for startups: every employee gets a scoped workspace — memory, files, keychain, crons, sandbox — in Slack and web, with Pi/OpenCode/Codex/Claude Code swappable underneath.
Same job — self-hosted platform where a team runs and supervises agents — different center of gravity: LobeHub is a chat UI for hiring/scheduling an agent workforce; qm is workspace infrastructure (scoped memory, keychains, sandboxes, crons) that meets people in Slack.
| lobehub | qm | |
|---|---|---|
| Stars | 81k | 7.2k |
| Forks | 16k | 758 |
| Language | TypeScript | TypeScript |
| License | NOASSERTION | MIT |
| Last activity | 4 days ago | 2 days ago |
| Topics | agents, orchestration | agents, orchestration |
| Curated connections | 3 | 3 |
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.
qm — the curator's take
The org-level answer to "give everyone at the company an agent": per-person and per-channel scopes keep memory, credentials and files isolated, and the harness underneath (Pi, OpenCode, Codex, Claude Code) is swappable, so the deployment isn't a bet on one vendor. Security postures (strict/auto/dangerous) plus a hard command-deny policy that survives even Dangerous mode are unusually well thought through. NOT for individuals — this is Postgres + sandbox + deployment-repo infrastructure; a solo dev should run a harness directly. Quirk: contributions are accepted as prose ADRs, not code PRs.