Zero-config data masking
Hoop redacts PII, PHI, and PCI data before it reaches Claude’s context, so the model can work with query results without seeing raw sensitive values.
Claude Code Gateway
Claude Code Gateway is Hoop.dev’s access-governance layer for Claude Code. It masks sensitive data, blocks dangerous commands, and records AI-assisted sessions so teams can use Claude Code with more control over what it can see and do.
Overview
Claude Code Gateway is Hoop.dev’s access-governance layer for Claude Code. It sits between the model and your infrastructure so teams can let Claude Code query systems, run commands, and interact with internal services without exposing raw secrets or allowing unsafe actions to pass through unchecked.
The product combines protocol-level masking, command blocking, human approval workflows, and session logging. The site positions it for teams that want Claude Code to keep operating with full context while policies determine what data it can see and what operations it can complete.
Features
Destructive command blocking
Dangerous commands such as DROP TABLE, DELETE without WHERE, and rm -rf are intercepted at the protocol layer and terminated before they execute.
Command-level review
Schema changes and other risky actions can be routed to Slack or Microsoft Teams for one-click approve/reject decisions, so execution pauses until a human responds.
Full session recording
Every command, response, approval, and rejection is recorded in a replayable audit trail for review and forensic use.
Broad protocol coverage
The gateway supports controlled access for databases, application runtimes, SSH, Kubernetes, web services, TCP, RDP, AWS SSM, and custom CLI tools.
Identity-based access governance
Pricing is organized around identities connecting through the gateway, with the site describing support for engineers, service accounts, and AI agents under one plan.
Use Cases
Safe internal data access
Let developers use Claude Code against internal databases or APIs while masking sensitive fields in query results and command output.
Approval-gated production changes
Require human review for schema changes, production writes, and other risky operations before the command executes.
Protocol-level guardrails
Stop destructive operations such as database drops or unsafe shell commands before they reach infrastructure.
Audit and forensic review
Keep a replayable record of AI-assisted sessions for compliance, audit, incident review, or post-incident analysis.
Centralized AI access control
Route Claude Code through a governed gateway so teams can control what it can see and do without changing Claude Code itself.
Pros and Cons
Pros
- Masks sensitive data before it reaches the model context, reducing exposure of raw PII, PHI, and PCI data.
- Blocks destructive commands at the protocol layer rather than relying only on post-execution review.
- Supports human approval for risky actions, including Slack-based one-click review.
- Keeps a detailed, replayable audit trail of commands and responses.
- Offers multiple deployment paths, including cloud signup and self-hosting options such as Docker, AWS, and Kubernetes.
Cons
- The public pages do not spell out all setup details for every environment, so some deployment and integration questions still require docs or a sales conversation.
- The page materials describe guardrails and audits well, but they provide limited concrete examples of full end-to-end team workflows beyond the homepage demos.
FAQ
What does Claude Code Gateway do?
Hoop sits between Claude Code and your infrastructure, so it can mask sensitive data, block destructive commands, and route risky actions for approval before execution.
How do you set it up?
The source shows setup paths for Docker, AWS CloudFormation, and Kubernetes, plus a cloud signup flow. It also says you can connect Claude Code to Hoop with one environment variable and no code changes.
How is Hoop priced?
The pricing page says one plan includes all features, with pricing based on the number of identities connecting through the gateway. It lists engineers, service accounts, and AI agents as identities.
Is it meant for team use?
Yes. The site describes access from humans, service accounts, and AI agents, and includes support for Slack and Microsoft Teams approvals, plus developer and self-hosted deployment options.
Quick Facts
- Category
- Developer Tool
- Product type
- Claude Code access gateway
- Primary users
- Engineers, DBAs, SREs, AI agents, and service accounts
- Deployment options
- Hoop Cloud, Docker, AWS, Kubernetes
- Approvals
- Slack and Microsoft Teams
- Website
- hoop.dev