Coordinate coding sessions, plans and code reviews across local and cloud work with HumanLayer.
HumanLayer is a collaborative workspace for coding agents and the people directing them. Its current product brings sessions, design documents, plans and code changes into shared tasks. The official site describes a coding IDE and cloud control plane rather than only the older concept of a human-approval SDK.
Structured workflows move through questions, research, design, structure, planning and implementation. Each stage creates an opportunity to correct an assumption before it becomes a large code change. Artifacts stay connected to the task, and comments can feed directly back into the agent’s work.
Teams can inspect session details, tool calls and diffs while using a higher-level task overview. Multi-repository workspaces and worktrees support changes that cross project boundaries. These features organise collaboration but do not replace architectural review or acceptance testing.
A local daemon runs sessions on a workstation, while remote daemons support background work elsewhere. The interface can be reached through web or desktop surfaces. HumanLayer advertises existing agent subscriptions or API keys, including Claude Code and Codex, rather than a separate bundled per-token model bill.
The Starter plan is free for small teams with a session allowance. Pro adds broader usage, while enterprise terms cover controls such as SSO and private deployment. The site says the product is not yet entirely open source, although its research-plan-implement framework is available separately.
Start with one bounded task and a team review checkpoint. Verify repository permissions and model billing, then inspect each planning artifact. Accept implementation only after tests and human review confirm that the result matches the agreed design.
