Understanding CVE-2026-63077 in On-Premises Build Servers
JetBrains has issued a critical security advisory warning administrators of an authentication bypass vulnerability affecting TeamCity On-Premises. Tracked as CVE-2026-63077, the flaw provides unauthenticated remote attackers with a direct mechanism to execute arbitrary operating system commands with the full privileges of the underlying TeamCity server process.
The security issue impacts all versions of TeamCity On-Premises. Any attacker capable of reaching an exposed TeamCity server over HTTPS can exploit the flaw without needing valid user credentials. In contrast, instances managed under TeamCity Cloud remain unaffected because JetBrains applied necessary server-side remediations prior to publishing the public advisory.
Security researchers privately reported the flaw to JetBrains on July 10, 2026. When JetBrains officially published its advisory on July 27, 2026, the company confirmed there was no evidence of active exploitation in the wild. However, public disclosure rapidly shrinks the window for mitigation. Continuous integration and continuous delivery (CI/CD) infrastructure represents a primary target for ransomware operators, nation-state threat groups, and financial extortionists. When a build server falls, the entire software supply chain breaks open.
Technical Mechanics: How the Agent Polling Protocol Is Abused
TeamCity relies on continuous, bi-directional communication between the central server and build agents deployed across corporate networks or cloud nodes. Agents repeatedly poll the central server over HTTPS to receive build assignments, fetch source repositories, stream build logs, and return compiled artifacts.
Under CVE-2026-63077, flaws within the handling of the agent polling protocol allow remote attackers to craft specialized HTTPS requests that circumvent server-side authentication gates. Rather than validating agent identities and session tokens strictly, the server process misinterprets incoming polling requests, allowing unauthorized commands to reach administrative execution contexts.
Daniel Gallo, Solutions Engineering Lead at JetBrains, highlighted the severe organizational risks associated with this vulnerability. Depending on process permissions, successful exploitation enables attackers to extract stored credentials, dump server configuration files, and inspect sensitive source code. Even more damaging, an attacker can modify build artifacts or inject malicious code directly into production deployments.
In modern continuous delivery, build servers hold administrative access across staging and production environments. Software organizations relying on automated deployment pipelines face complete operational disruption if an unauthenticated adversary compromises the build engine.
AI Cybersecurity Threats and Build Pipeline Risks in 2026
The emergence of CVE-2026-63077 underscores how AI cybersecurity threats in 2026 have altered the speed of vulnerability discovery and exploitation. Threat actors routinely deploy automated reconnaissance engines and machine learning scanners to map public IPv4 and IPv6 subnets for exposed management endpoints, REST APIs, and build tools.
When automated tools discover exposed TeamCity servers, AI cybersecurity threats transition into swift weaponization. Sophisticated threat actors utilize AI agent security frameworks to test payload variations against agent polling endpoints within hours of security announcements. This rapid automated probing leaves unpatched infrastructure highly vulnerable, as demonstrated in recent lessons from autonomous breaches.
Enterprise security research from organizations such as IBM highlights that CI/CD pipelines remain high-yield vectors for supply chain compromises. Penetrating a single development server allows attackers to bypass endpoint protections downstream. As engineering teams integrate agentic systems into build scripts, securing foundational build nodes becomes indispensable for complete defense. You cannot protect agentic applications if the server compiling their code is compromised.
Our platform analysis on securing agent infrastructure demonstrates why organizations must move beyond passive firewalls to continuous patch management and aggressive zero-trust network segmentation.
Step-by-Step Remediation Tutorial and Security Patching
To eliminate exposure to CVE-2026-63077, JetBrains advises organizations running TeamCity On-Premises to upgrade immediately. The vendor addressed the vulnerability in TeamCity maintenance versions 2025.11.7 and 2026.1.3. Upgrading to these releases provides complete resolution.
For organizations unable to execute an immediate major version upgrade, JetBrains released a standalone security patch plugin. This section serves as an operational tutorial for securing legacy installations across corporate networks:
- Verify Server Eligibility: The standalone security patch plugin supports TeamCity On-Premises versions 2017.1 through recent releases.
- Automated Plugin Download (Versions 2024.03+): Installations running TeamCity 2024.03 or newer automatically download available security patch plugins in the background and notify administrators within the console. Admins need only approve the installation prompt.
- Manual Plugin Installation (Versions 2017.1 to 2023.11): Download the official patch plugin from JetBrains. Navigate to
Administration -> Pluginsin the TeamCity console, upload the plugin ZIP archive, and enable it. - Enforce Server Restarts Where Mandatory: For installations running TeamCity 2017.1 through 2018.1, enabling the plugin is insufficient on its own. A full server process restart is required for security changes to take effect. Versions 2018.2 and newer apply the plugin dynamically without requiring a server reboot.
Following patch application, security teams should inspect active server sessions and agent connections to verify that all build nodes re-authenticate successfully. Detailed reporting on this incident was originally published by BleepingComputer.
Defenses and Best Practices for Agentic Infrastructure
Beyond applying emergency software patches, JetBrains urges administrators to implement systemic defenses to protect build environments against active vulnerability exploitation and emerging AI cybersecurity threats.
Adhering to cybersecurity best practices recommended by the Cybersecurity and Infrastructure Security Agency (CISA) and enterprise operational guidelines from IBM, organizations should enforce strict access controls and identity gateways for autonomous AI agents around software delivery pipelines:
- Restrict Network Exposure: Never leave TeamCity web portals or REST API endpoints directly accessible from the public internet. Require enterprise VPN connectivity, Zero Trust Network Access (ZTNA), or strict IP firewall rules for all server traffic.
- Isolate Build Agents: Run build agents inside isolated network segments or ephemeral containers. Ensure agents cannot reach internal core databases or unrelated corporate assets.
- Perform Credential Hygiene: Rotate stored access tokens, cloud service keys, and repository credentials immediately if server compromise is suspected. Regular credential rotation limits lateral movement during zero-day events.
- Monitor Polling Anomalies: Log all incoming agent polling requests and API calls. Establish alert thresholds for unexpected IP addresses or unusual command execution patterns originating from build tasks.
Build servers are the operational backbone of modern engineering teams. Securing them requires proactive maintenance, rapid patch installation, and continuous architectural isolation. By pairing immediate updates with robust defense practices, enterprises can safeguard their CI/CD pipelines against automated attacks throughout 2026.
