Chainstack breach exposes account data; customers urged to rotate node credentials

The breach occurred when an attacker exploited a test Grafana instance deployed in July that still used default administrator credentials and lacked additional authentication. This allowed the attacker to access internal services and eventually reach customer databases containing organization names, account holder names, members' email addresses and password hashes. While Chainstack maintains no evidence that node credentials were read, the company is treating them as potentially exposed.

The attacker's path into Chainstack's systems began through a development cluster's Grafana instance, which had no destination allowlist and unrestricted network policies. Over approximately two hours, the attacker connected to dozens of internal services, including an MCP tool server with cluster-admin permissions that exposed Kubernetes operations to AI agents.

Attack progression

On October 29, the attacker compromised an automation platform by rewriting a workflow in its database to execute shell commands, obtaining environment variables and service-account tokens. Using a stolen VPN peer private key on October 31, the attacker gained access to the database network, triggering Chainstack's investigation when it disrupted database replication.

The attacker used a compromised Kubernetes token to query PostgreSQL's password-hash view from a database connection-pooler pod.

Customer response required

Chainstack is requiring customers to rotate node credentials, which involves generating new credentials, migrating applications while both sets remain functional, then permanently deactivating the old set. The process affects HTTPS and WSS endpoint URLs, password-protected access credentials, and gRPC endpoint authentication values.

Customers must also change passwords on any other services where they reused their Chainstack password and regenerate platform API keys. Chainstack is warning customers about potential phishing attempts, emphasizing it will never request passwords, tokens or keys directly.

Security gaps identified

The incident revealed multiple security failures, including default credentials, missing authentication layers, excessive permissions and long-lived credentials. Chainstack attributed the attack tooling to ARTEX, an open-source framework for language-model agents, but found no evidence of exploited software vulnerabilities.

Despite the breach, Chainstack confirmed that customer-facing services remained operational throughout the incident, with no changes to customer nodes, console or API. The company also verified that full payment-card numbers and security codes were not exposed as it doesn't store them.

Chainstack completed its investigation on November 2, 2026, and continues to monitor for any remaining suspicious activity while implementing additional security measures to prevent similar incidents.