Run Claude Code and Codex as autonomous agents, in your own cloud

The self-hosted alternative to Claude Managed Agents: headless, vendor-independent, on your Kubernetes. Turns can run on your Claude or ChatGPT plan, or on open-weight models, with no per-token API bill from a model vendor.

Self-deployed in your Kubernetes

Air-gapped ready

SOC 2 Type II

GDPR

The managed-agents experience, in your own Kubernetes

Claude Managed Agents runs Claudeโ€™s harness for you on Anthropicโ€™s infrastructure. xpander runs the harnesses your teams already use, on any model, inside your cluster, and governs every call they make.

Claude Managed Agents

xpander

Harness

Harness

Claudeโ€™s own agent harness

Claude Code, Codex, and OpenCode, running headless

Models

Models

Claude

Anthropic, OpenAI, Bedrock, and open-weight models

Where it runs

Where it runs

Anthropicโ€™s cloud. Sandboxes can be self-hosted, but session history is stored by Anthropic.

Your Kubernetes: VPC, on-premises, or air-gapped. With an open-weight model, nothing leaves your perimeter.

Cost

Cost

API tokens at list price, plus $0.08 per session-hour of runtime

Turns can run on your Claude or ChatGPT plan, or on open-weight models on your own GPUs

Lock-in

Lock-in

One model vendor

Switch harness and model per conversation

Works with the AI your teams already use

Teams keep the tools they like. IT sets the rules once, for all of them.

Runs coding agents

  • Claude Code

  • Codex

  • OpenCode

  • Grok Build

  • pi

  • Your own

Works with models

  • Claude

  • Gemini

  • Llama

  • GPT

  • Amazon Bedrock

  • Qwen

  • DeepSeek

  • Z.ai

  • Mistral

  • Open-weight models

  • Your fine-tuned models

Deploys to

  • AWS

  • GCP

  • Azure

  • Your VPC

  • Air-gapped on-prem

Lives in

  • Slack

  • Microsoft Teams

  • WhatsApp

  • Telegram

  • Claude Code

  • Desktop app

  • Any editor

  • Whatever you adopt next

Built for the most security-conscious teams

Certified and audited

SOC 2 Type II certified and GDPR compliant, with independently audited controls so your security review starts from a known baseline.

Runs where your data lives

Your VPC, your data center, or fully air-gapped. Nothing leaves your environment unless you say so.

Every action is logged

Agents act as the person who asked, credentials come from your vault at call time, and every run is logged and attributable.

See how it would run in your environment