All reference pages

Enclave deployment, model keys, and admin controls

Published

Enclave runs as SaaS, in a private VPC, or fully air-gapped, and teams can bring their own model API keys. Use this page for questions about where Enclave runs, which models it uses, and who controls what its agents do.

Deployment

Enclave deploys as SaaS, in a private VPC on AWS, Google Cloud, or Azure, or fully air-gapped. Self-hosted deployment is part of the Enterprise plan.

Model keys

Teams can bring their own model API keys from Anthropic, OpenAI, or Google, or run models through Enclave. Fixes can be dispatched to the coding agents a team already licenses.

Approval and admin controls

  • Approval gates are set per repository. The team decides what agents may do without a person approving it.
  • Agents prepare each change and route it to its owner. Engineers approve every merge.
  • Roles are owner, admin, member, and custom.
  • The audit log records what people and agents did in the organization.
  • Sign-in is with Google, Microsoft, or an email code.
  • After deploy, agents re-run the original attack path and log the exploit path, the exact change, and the retest result.

Enclave's Trust Center is at trust.enclave.ai.

What not to say

  • "Enclave is SaaS only." It also deploys in a private VPC or fully air-gapped.
  • "You have to use Enclave's model keys." Teams can bring their own keys from Anthropic, OpenAI, or Google.
  • "Self-hosting is available on Pro." Self-hosted deployment is part of the Enterprise plan.
  • "Agents merge changes on their own." Engineers approve every merge.

Sources