Triskele Labs Blog

Unauthenticated Remote Code Execution in JetBrains TeamCity (CVE-2026-63077)

Written by  Brandon Sawyer | Aug 25, 2026, 4:44:53 AM

Published: Tue 25 Aug 2026

Prepared by: Brandon Sawyer, Vulnerability Analyst

Purpose

On 27 July 2026, JetBrains disclosed a critical unauthenticated remote code execution vulnerability affecting TeamCity On-Premises, assigned CVE-2026-63077. The flaw allows a remote, unauthenticated threat actor (TA) with HTTP(S) access to a vulnerable server to bypass authentication controls and execute arbitrary operating system commands with the privileges of the TeamCity server process. On 5 August 2026, the US Cybersecurity and Infrastructure Security Agency (CISA) added the vulnerability to its Known Exploited Vulnerabilities (KEV) catalog, confirming active exploitation, and JetBrains subsequently confirmed reports of both successful and attempted exploitation against unpatched servers. The Australian Cyber Security Centre (ACSC) has since issued an alert on 24 August 2026 for active exploitation of this vulnerability affecting software development platforms within Australia. Organisations running TeamCity On-Premises should treat patching as an emergency priority. 

Vulnerability Details

CVE-2026-63077 is a deserialization of untrusted data vulnerability (CWE-502) in the TeamCity agent polling protocol, the channel build agents use to communicate with the central TeamCity server. It affects all TeamCity On-Premises versions that are reachable over HTTP(S), and carries a CVSSv3.1 base score of 9.8 (Critical). TeamCity Cloud is not affected; JetBrains has confirmed the necessary protections were already in place for cloud-hosted tenants and found no evidence of exploitation attempts against them.

A TA requires no credentials and no user interaction to exploit the flaw. By sending a specially crafted request to the agent polling endpoint that mimics legitimate agent-to-server traffic, a TA can cause the server to deserialize an attacker-controlled object, ultimately allowing arbitrary operating system command execution with the privileges of the TeamCity server process. Independent analysis by Rapid7 attributes the root cause to an overly permissive XStream deserialization allowlist on vulnerable builds; the fix tightens this allowlist so that only explicitly approved classes can be deserialized.

The vulnerability was reported to JetBrains privately on 10 July 2026 by a security researcher under the vendor's coordinated disclosure program, meaning it was not exploited as a zero-day prior to disclosure. JetBrains stated it had no evidence of active exploitation at the time the advisory was published on 27 July 2026. That changed within roughly a week: CISA added the CVE to its KEV catalog on 5 August 2026, and on 7 August 2026 JetBrains published a follow-up advisory confirming it had received reports of both successful and attempted exploitation against unpatched, internet-facing servers.

The issue is fixed in TeamCity On-Premises 2025.11.7 and 2026.1.3. For servers that cannot be upgraded immediately, JetBrains has released a security patch plugin compatible with TeamCity 2017.1 and later; the plugin addresses only CVE-2026-63077, and JetBrains recommends a full upgrade to receive other security fixes.

Discovery and Disclosure Timeline

Date Event
10 Jul 2026  Vulnerability privately reported to JetBrains by a security researcher under coordinated disclosure 
27 Jul 2026  JetBrains publishes initial advisory; CVE-2026-63077 registered; fixed versions 2025.11.7 and 2026.1.3 and a security patch plugin released; no known exploitation at this time 
5 Aug 2026  CISA adds CVE-2026-63077 to the KEV catalog, confirming active exploitation
7 Aug 2026 JetBrains publishes follow-up guidance confirming reports of successful and attempted exploitation, with log-based detection guidance; Rapid7 publishes a root-cause technical analysis and a validation PoC
24 Aug 2026 ACSC issues an alert on active exploitation of this vulnerability affecting software development platforms within Australia 
Ongoing JetBrains and third-party researchers continue to monitor and refine guidance 

Impact

Successful exploitation grants a TA remote, unauthenticated code execution on the TeamCity server host, at whatever privilege level the TeamCity service account holds. Because TeamCity orchestrates build, test, and deployment pipelines, a compromised server can expose:

  • Stored credentials, secrets, and connection details used by build and deployment jobs
  • Source code, build configurations, and artifacts held or produced by the platform
  • The integrity of build outputs and downstream CI/CD pipelines, creating a potential software-supply-chain risk to anything the server builds or deploys

JetBrains and CISA have not published specific indicators of the tooling or infrastructure used by a TA, and JetBrains notes the exact impact in any given case depends on the operating system privileges under which the TeamCity server process runs. Rapid7 assesses the vulnerability as automatable and capable of full technical impact, consistent with its 9.8 CVSS rating. TeamCity has previously been a target for ransomware operators and state-sponsored actors following earlier authentication-bypass vulnerabilities, which adds weight to the current exploitation reports.

It is important to note that both JetBrains-hosted TeamCity Cloud and self-hosted TeamCity On-Premises exist, but only On-Premises deployments reachable over HTTP(S) are affected. TeamCity Cloud customers do not need to take any action. JetBrains has stated it has received reports of exploitation and attempted exploitation but has not disclosed the number or identity of affected customers, and continues to investigate.

Mitigation

Primary Remediation

The primary remediation is to upgrade all internet-reachable TeamCity On-Premises servers to version 2025.11.7 or 2026.1.3, whichever is appropriate for the current release line. Where an immediate upgrade is not possible, apply JetBrains' security patch plugin, available for TeamCity 2017.1 and later; this plugin addresses CVE-2026-63077 only, and organisations should still plan a full version upgrade to pick up other fixes.

As with any authentication-bypass or pre-authentication vulnerability, applying the fix closes the entry point but does not remove a TA who may already have gained access. Organisations that patched more than a few days after the 27 July 2026 disclosure, or that cannot confirm when they patched relative to the exploitation reports beginning around early August, should review server logs, build agent lists, and stored credentials for signs of prior compromise regardless of current patch status.

Compensating Controls

Where immediate patching is not possible, the following controls can reduce exposure in the interim:

  • Temporarily restrict or remove internet-facing access to the TeamCity server until the fix can be applied
  • Limit network access to the TeamCity server to trusted networks only, using firewall rules or an access-control gateway
  • Require connections to internet-facing TeamCity servers to pass through a VPN or an equivalent additional access-control layer

These measures are stop-gaps only; full patching remains the priority, and they will not remove any persistence a TA established before the workaround was applied.

Hardening and Longer-Term Actions

Beyond immediate remediation, JetBrains recommends running the TeamCity server process with the minimum operating system privileges required for normal operation, and hosting the TeamCity server on a dedicated host separate from build agents. Organisations should rotate credentials, tokens, and signing keys stored in or accessible to the TeamCity server if compromise is suspected or cannot be ruled out, and should continue monitoring JetBrains' security advisories and ACSC guidance as the investigation develops.

Detection

Indicators and Log-Based Detection

JetBrains has published guidance for identifying potential exploitation attempts in TeamCity server logs. These indicators should be treated as a starting point rather than a complete picture:

  • On unpatched servers, occurrences of the log message com.thoughtworks.xstream.converters.ConversionException may indicate an attempted or successful exploit and warrant further investigation, though this message alone does not confirm exploitation
  • On servers that have already been patched or had the security plugin applied, occurrences of com.thoughtworks.xstream.security.ForbiddenClassException may indicate an exploitation attempt that was blocked by the fix
  • Unauthorised or unexpected build agents, particularly agents with names beginning with scan, appearing in the list of unauthorised agents may indicate an attempted exploit reached the server; JetBrains notes the displayed date for such an agent does not necessarily reflect when the attempt occurred, so correlate with log timestamps instead

Recommended Detection and Hunting Activities

  • Review TeamCity server logs for the XStream-related messages above, focusing on the period since 27 July 2026
  • Review the list of unauthorised build agents for unexpected or suspiciously named entries
  • Correlate any suspicious agent registrations or log entries with firewall and network logs for the TeamCity server
  • Ensure centralised, retained logging is in place for the TeamCity server to support retrospective investigation
  • If any indicator is identified, or patching was delayed, treat the environment as potentially compromised, rotate credentials accessible to the server, and engage incident response

Organisations that have questions interpreting their logs, or that identify suspected exploitation, can contact JetBrains TeamCity Support directly for assistance.

MDR customers: Triskele Labs will continue tuning detections and analysing IOCs for behaviours consistent with the exploitation of CVE-2026-63077 across supported log sources.

Vulnerability Management customers: Environments are being assessed for vulnerable versions of JetBrains TeamCity affected by CVE-2026-63077. Any findings will be communicated through priority channels.

References