About HumanLayer
HumanLayer is a multiplayer IDE and cloud for running coding agents as a team. A daemon on each host runs Claude Code, Codex, OpenCode or Pi sessions: a local daemon runs many sessions in parallel on a laptop, cloud daemons handle long-running work, and everything syncs through the HumanLayer API to the desktop app, the web app or a phone. Tasks group sessions, plan artifacts, worktrees and diffs so humans and agents can review the work together using Research-Plan-Implement workflows.
In October 2026 HumanLayer open-sourced its core under the Apache 2.0 license. The GitHub repository holds the desktop apps (Electron and Tauri), the web app, the daemon shipped in the humanlayer CLI, the API, the agent plugins and the Research-Plan-Implement skills, and HumanLayer Cloud runs the same code. According to the README, a build from source sends no telemetry because error tracking and analytics stay off unless configured. Self-hosting guides are still in progress, and the sample Docker Compose file uses default passwords that must be changed before deployment.
Best-fit usage: teams coordinate multiple agent sessions, plans, artifacts, and review steps across complex codebases. The tool page should guide readers toward that workflow while making the major limitation clear: reducing the product to old human-in-the-loop approval language. For aicoolies readers, that distinction matters because HumanLayer should be judged on verified buyer fit, not on copied launch phrasing or assumptions that may have drifted since the last CMS update.
Procurement note: HumanLayer Cloud's Starter plan is free for up to 3 team members and 200 sessions per month, Pro costs $100 per user per month, and Enterprise adds SSO/SAML, audit logs, volume pricing, custom terms and on-prem or private VPC deployment. HumanLayer states SOC 2 Type II compliance, with ZDR and DPA agreements available on enterprise tiers. Model usage runs on your own Claude, Codex or other subscriptions and API keys, so compare total cost as seats plus model spend, and verify data handling and exit paths before production use.
