Protocol-aware inspection
The proxy interprets statements and calls carried over supported protocols, rather than treating traffic as opaque bytes. The documentation names PostgreSQL, ClickHouse, SQL Server, and HTTP, among others.
A protocol-aware sidecar gateway that controls AI agent and infrastructure traffic.
hoop.dev is an open-source, protocol-aware reverse proxy for controlling interactions between clients, AI agents, and infrastructure resources. Deployed as a sidecar beside a database, cluster, or other supported resource, it inspects traffic at the protocol layer and applies controls before requests or responses reach their destination. Its core purpose is to govern activity after authentication: sensitive response data can be masked, prohibited commands refused, and selected actions held for review. Existing clients and credentials can remain in place, with no application code changes required. An optional Enterprise control plane manages rules across a fleet of sidecars.
The proxy interprets statements and calls carried over supported protocols, rather than treating traffic as opaque bytes. The documentation names PostgreSQL, ClickHouse, SQL Server, and HTTP, among others.
Sensitive values such as emails can be rewritten in memory before the response reaches a client. The documented masking path changes responses, not requests.
Ordered deny rules can refuse statements or commands that should not run, with a configured message explaining the refusal.
For agentic access, an analyzer classifies statements and risk, then can block, allow, request review, apply a guardrail, or mask a response. The open-source offering lists block and allow actions; Enterprise includes the full Session Analyzer, including warn and defer.
The sidecar records activity passing through it, providing a session record alongside its inline controls.
The Enterprise control plane manages rules across sidecars, removes per-sidecar rule limits, and shows which rules are live and how they resolved.
Route an agent's database access through a sidecar so it can query needed records while sensitive response fields are masked before the agent receives them.
Evaluate writes before they reach a database; allow lower-risk changes to proceed and route selected higher-risk actions to a human for approval.
Apply deny rules to operations such as destructive database statements or cluster actions so matching commands are refused before execution.
For teams operating multiple protected resources, use the Enterprise control plane to publish rules centrally and inspect their status across sidecars.
The sidecar is described as a no-code proxy. Existing clients can stay the same, and the product is designed to work without application changes.
It runs beside the resource it protects. Installation options include the hoop CLI, container images, and a Kubernetes Helm chart.
The MIT-licensed, self-hosted sidecar includes data masking, guardrails, Session Analyzer, and session audit. Each sidecar is configured from a file and is limited to one rule of each kind.
Enterprise adds a control plane for managing rules across sidecars, removes rule limits, includes the full Session Analyzer with warn and defer, and offers support and SLAs. Deployment is in the customer's infrastructure.
The pricing page says Enterprise pricing is provided based on the customer's environment. It counts engineers, AI agents, and service accounts as identities; no specific price is published.
Os dados de tráfego são apenas para referência.
theagentrouter.ai
Agent Router is an open-source Go project for managing unified access to generative AI services through Envoy Gateway. It is intended for teams that want a gateway-based control point for AI service traffic and related agent or MCP workloads.
glama.ai
Glama is a registry and infrastructure platform for discovering, evaluating, and connecting to Model Context Protocol (MCP) servers, connectors, clients, and tools. It supports browsing open-source and remote MCP entries, with hosted access and gateway features described on the site.
oceanprotocol.com
Open-source infrastructure for developers to monetize AI models and data with privacy controls.
www.cohesor.com
Cohesor is an AI gateway and control plane for routing model requests, compressing agent context, governing spend, and brokering MCP tool calls. It is designed for teams running internal or customer-facing AI agents.
www.docker.com
Docker MCP Enterprise Gateway governs how MCP-compatible clients access approved servers and tools. It applies identity-aware policy, supplies credentials from an approved secret store, and records tool-call decisions for enterprise teams.
www.edgee.ai
Edgee is an agent gateway for coding teams that reduces token usage, routes requests across models, and provides usage visibility. Its Compression V2 works as a drop-in CLI layer for Claude Code, Codex, OpenCode, Cursor, and other supported agents.