Before engineers connect company code, settle where inference goes, what Warp collects, and how secrets are masked. Then set sign-in controls and define what the agent can run. Each check names the plan that carries the control.
When you need this
- Security has to approve a terminal that sends commands and output to model providers.
- Engineers installed Warp on their own and you need one enforced configuration.
- You are deciding between Business and Enterprise and need to know which controls each carries.
Security review checklist
Zero data retention scope
Which requests are covered, and which are not?
SetWarp's agreements with Anthropic, OpenAI, Google, xAI, and Fireworks AI cover Warp-provided inference on every plan. Requests through your own keys, a connected subscription, or a custom endpoint follow that provider's terms. Some models have provider retention requirements that exclude them from zero data retention. On Enterprise, those models stay off until an admin enables them.
Telemetry and data collection
Is collection on or off by default for your plan?
SetOn Business and Enterprise, data collection is off by default and admins enforce it for the team. On Build and Max teams it is on by default, and each user can turn off Help improve Warp in Settings. Warp publishes every telemetry event and shows live requests in its network log.
Secret Redaction
Are credentials masked before a request leaves the machine?
SetSecret Redaction is off by default. Turn it on with Warp's recommended patterns plus your own, and on Enterprise enforce it from the Admin Panel so no one can switch it off.
Identity
How do people sign in and lose access?
SetSAML SSO from Business, set up with Warp through WorkOS, with Okta, Microsoft Entra ID, Google Workspace, OneLogin, or any SAML or OIDC provider. Admins can require SSO. Provisioning is just-in-time with domain auto-join, and removing a user in the identity provider blocks new sign-ins without ending existing sessions.
Agent permissions
What can the agent run without asking?
SetAdmin-enforced autonomy, command denylists, and model policies are Enterprise features. On lower plans each engineer sets their own profile. Keep the denylist bypass for auto-approve turned off.
Model routing and hosting
Where do model calls and cloud runs execute?
SetEnterprise BYOLLM routes eligible requests through Claude on AWS Bedrock or Gemini and Claude on Google Vertex AI, with no stored keys. Vertex does not serve cloud agents yet. Warp-hosted cloud agents run in ephemeral Docker containers on Google Cloud. Self-hosted execution keeps clones and secrets on your machines, and transcripts and inference still route through Warp's backend.
The client itself
Can your security team read the code?
SetThe client is open source under AGPL v3 at github.com/warpdotdev/warp, published on April 27, 2026. Warp holds a SOC 2 Type II attestation, with the report available through its Trust Center.
Questions to settle before rollout
- Warp's legal security page offers zero data retention on Enterprise, while its pricing and docs apply it to every plan. Which does the contract state?
- Which models outside zero data retention are enabled for our team, and who approved them?
- How long are shared sessions kept? Warp's documentation says about one week, with no way to delete one sooner short of deleting the account.
- Does a security incident on a self-hosted run expose transcripts held by Warp?
Red flags
- A Build or Max team handling regulated code with data collection left on by default.
- Personal API keys in use where the company has no agreement with that provider.
- Secret Redaction off on machines that hold production credentials.
- Deprovisioning relied on to end sessions immediately.
Products in this piece
Sources
- Official documentation
- 15
- Warp docs: Security overview
- Warp docs: Privacy and data control
- Warp docs: Warp network log
- Warp docs: Secret Redaction
- Warp docs: Single Sign-On (SSO)
- Warp docs: Admin Panel for teams
- Warp docs: Bring Your Own LLM
- Warp docs: Team-managed LLM API keys and endpoints
- Warp docs: Architecture and deployment
- Warp docs: Security and networking for self-hosted agents
- Warp docs: Agent Session Sharing
- Warp: Pricing
- Warp: Security
- Warp on GitHub: warpdotdev/warp
- Warp docs: Changelog 2026