How to Configure Neo4j Proxy
You set up a neo4j proxy by routing HTTP and Bolt traffic through a forwarder. You also update the advertised host address in the graph database settings. Most teams start by sharing static database keys and opening network ports to developer laptops. Broad roles let any logged-in user run large graph traversals or delete production nodes. How a neo4j proxy handles graph traffic A graph database uses two main channels for client traffic. The web browser interface uses standard HTTP or HTTPS c
Your coding agents run with your access.
Claude Code reaches prod with a developer's credentials, and approval prompts get clicked through. hoop.dev checks every query, masks sensitive data, and holds risky changes for a human.
You set up a neo4j proxy by routing HTTP and Bolt traffic through a forwarder. You also update the advertised host address in the graph database settings. Most teams start by sharing static database keys and opening network ports to developer laptops. Broad roles let any logged-in user run large graph traversals or delete production nodes.
How a neo4j proxy handles graph traffic
A graph database uses two main channels for client traffic. The web browser interface uses standard HTTP or HTTPS calls. Database drivers and apps use the Bolt protocol over TCP sockets. When you place a load balancer in front of the cluster, both channels need routing rules.
HTTP traffic works with standard web forwarders. The forwarder reads the host header and sends requests to the right backend node. It also adds headers that tell the database the original client address. The database reads these headers to build clean browser links.
Bolt traffic is different because it uses persistent binary TCP sockets. After a client opens a Bolt session, the database sends back a routing table. This table lists the internal addresses of core and read replica nodes. If the database sends private IP addresses, external clients fail to reach the cluster.
To fix this, admins change the advertised address in the database settings. Each cluster node must report the public host name of the network router. The router then listens on distinct ports and maps each port to a specific backend node.
Why native roles and network routing fall short
Network routing and native database roles solve the first layer of access control. They keep raw cluster nodes off the public internet. Native roles let admins split users into readers, editors, and system owners. You can also bind these roles to an external directory service.
Still, network routers only pass packets from point A to point B. They do not inspect the Cypher queries running inside the binary Bolt stream. Once a user logs in with a valid role, the router stops checking their actions.
Database Proxy (ProxySQL, PgBouncer) + End-to-End Encryption: architecture patterns and best practices
Continue reading? Get the full guide.
Free. No spam. Unsubscribe anytime.
Native roles also stay broad in day-to-day work. An engineer who fixes a bug often gets full write access to the whole graph. That engineer can run an unbounded query that locks memory or dumps all user nodes.
Also, native query logs live on the database servers and require extra setup to export. When ten developers share one admin user account, the logs show the same account name. Teams cannot tie a bad Cypher delete statement back to a single person. Sensitive fields like emails, phone numbers, card numbers, and national IDs also return in plain text.
How to plan ports and TLS for graph clusters
Good routing starts with a clear port map and clean TLS certificates. Each node in a causal cluster needs its own reachable endpoint for Bolt routing. If three core nodes sit behind one router, you assign three public ports on the router.
TLS termination is the next big choice for the ops team. You can end TLS at the router or pass encrypted traffic straight to the database. Ending TLS at the router makes certificate updates simple because you manage keys in one place. Passing TLS through to the database keeps traffic encrypted across the internal network too.
Health checks keep the routing table accurate during node restarts. The router should ping the HTTP status endpoint of each graph node every few seconds. When a node drops out of the cluster, the router stops sending new sessions to it. This setup keeps apps stable, though it still leaves user keys and query data unmanaged.
Govern statements and responses across databases and APIs
Good config and scoped roles are needed setup. They do not check single queries in flight or mask sensitive fields in returned rows. Across Postgres, MySQL, SQL Server, MongoDB, ClickHouse, Kubernetes APIs, SSH, and internal HTTP endpoints, hoop.dev sits inline in the protocol beside each supported service as an open-source proxy, a sidecar for your database or API. For every query or command an AI agent or engineer runs against production, hoop.dev decides to pass, mask, block, or hold it for human approval.
First, hoop.dev guardrails block banned queries before they run or hold risky ones for review, and hoop.dev data masking rewrites matched sensitive values in responses in memory. Furthermore, hoop.dev holds SOC 2 Type II and is open source on GitHub.
Frequently asked questions
Why do Bolt clients fail after the initial login through a load balancer?
The database returns internal private IP addresses in its Bolt routing table by default. You must set the advertised address on each node to match the public router host name.
Can a standard web router balance Bolt traffic across cluster nodes?
A web router can forward TCP streams, but Bolt drivers manage their own routing. The router must expose each node on a distinct port so the driver can pick leaders or replicas.
How do teams protect sensitive graph properties during developer troubleshooting?
Native roles restrict node labels or properties, which often breaks app queries during debugging. Data masking rewrites matched fields like emails or card numbers in results while keeping queries working.
Related guides
In this post