Nuphos is changing at its foundation: we are no longer asking you to adopt a Nuphos-built AI agent.
Instead, Nuphos is becoming the DevOps platform for the agents your team already uses. Our integration boundary is the terminal coding agents supported by OpenAB. Bring Claude Code today—with other agents following the same runtime path as their adapters are enabled—and let Nuphos provide the shared infrastructure context, credentials, tools, memory, approvals, and collaboration layer around it.
Your agent, inside your workspace
Workspace administrators can now connect Claude Code with a token generated
by claude setup-token. Nuphos prepares an isolated
OpenAB runtime for that workspace and
uses the team's Claude account for every conversation.
This means the model access is yours. Claude Code runs with the subscription and identity you already use, and its inference does not consume Nuphos agent credits.
The first agent enabled in Nuphos is Claude Code. The integration boundary is the terminal coding agents supported by OpenAB, which already bridges agents such as Claude Code, Codex, Gemini CLI, Kiro CLI, and more through their native harnesses. It is not limited to agents that implement Nuphos's remote MCP interface.
Nuphos supplies the DevOps context
Changing the runtime did not mean giving up the parts that make an agent useful and safe in production. Your connected agent gets Nuphos capabilities scoped to the current team and conversation:
- short-lived credentials for the cloud accounts selected in the session;
- cloud, Kubernetes, observability, database, and Nuphos API skills;
- team memory and reusable team-authored skills;
- structured plans, risk and cost context, and human approval before changes;
- persistent transcripts, pause and resume, and mid-turn steering; and
- the same conversations from the desktop app, Slack, iMessage, triggers, and MCP.
Nuphos provisions a runtime per workspace, keeps conversations attached across restarts, and streams Claude Code's thinking and tool activity into the same shared transcript your team sees in the app.
Why we made the change
Developers already have an agent they trust, customize, and use all day. A separate DevOps agent creates another context silo and asks teams to relearn a workflow they already have.
We think the better product is the layer that makes your agent production ready: one place for infrastructure access, operational knowledge, team collaboration, and human control. You keep the agent; Nuphos gives it a shared DevOps workspace.
Will this get my Claude Code account blocked?
Nuphos is designed to avoid that restriction: it does not turn a Claude subscription into a raw API or replace Claude Code with a third-party harness.
OpenAB works like a bridge between a terminal coding agent and a conversation surface such as Slack, Discord, or Telegram. Nuphos occupies the same position as one of those conversation surfaces: the prompt travels through the bridge, but Claude Code itself still runs the request with Anthropic's own CLI, authentication, tool loop, and harness.
That is different from the OpenClaw pattern Anthropic restricted. That policy prevents third-party harnesses from using Claude subscription authentication to bypass Claude Code's harness and consume the subscription like an API. Nuphos keeps each vendor's native harness intact. It adds Skills, MCP servers, team memory, credentials, and DevOps context around the agent instead of replacing the agent's execution layer.
Providers ultimately control their own terms and enforcement, but Nuphos's architecture does not depend on the subscription-auth bypass that policy was designed to stop.
Connect Claude Code
Update to the latest Nuphos desktop app, then open Settings → Workspace →
Agent. Run claude setup-token on a machine where Claude Code is signed in,
bind the generated token, and wait for the workspace runtime to show
Online.
For the complete setup and security model, see the agent runtime documentation.