AWS Labs Timestream for InfluxDB MCP Server Vulnerable to API Token Leak via Prompt Injection (CVE-2026-126681)
Disclosed on November 16, 2026, CVE-2026-126681 is a prompt injection vulnerability in the official AWS Labs Timestream for InfluxDB MCP Server (v0.0.12 to v0.0
A significant security vulnerability has been discovered in the official Timestream for InfluxDB MCP Server developed by AWS Labs, an open-source and security research organization within Amazon Web Services. Disclosed on November 16, 2026, and tracked as CVE-2026-126681, this flaw allows an attacker to exfiltrate database API tokens to an external endpoint via prompt injection during interactions with AI assistants.

Image source: https://i10x.ai/news/cve-2026-126681-aws-labs-timestream-influxdb-mcp-server-prompt-injection
This incident underscores the security risks inherent in AI agent integrations and Model Context Protocol (MCP)—an open standard designed to enable secure communication between large language models and external data sources or developer tools. Because the attack requires no authentication or special privileges, developers and organizations operating affected servers must take immediate remediation steps.
Overview and Attack Mechanism (CVE-2026-126681)
The affected package is awslabs.timestream-for-influxdb-mcp-server, spanning versions v0.0.12 through v0.0.18. It has been assigned a CVSS v3.1 score of 5.5 (Medium severity).
- Root Cause: The vulnerability stems from improper restriction of intended endpoints in the server's connection configuration handling (CWE-923), failing to validate where connections are directed.
- Attack Vector: A remote unauthenticated attacker can inject a crafted InfluxDB URL into the MCP client context using prompt injection techniques.
- Result: When the AI assistant processes the injected prompt, the server attempts to connect to the attacker-controlled URL, transmitting its configured InfluxDB API token directly to the external endpoint.
If exploited, the confidentiality impact is high, as leaked API tokens grant direct access to underlying database operations based on the permissions assigned to the token.
Upgrading and Emergency Remediation
AWS Labs published a security advisory addressing the issue in version v0.0.19.
- Applying the Patch: Users must immediately upgrade the MCP server to version 0.0.19 or later. Installation can be performed via PyPI using the command:
pip install awslabs.timestream-for-influxdb-mcp-server==0.0.19. - Rotating API Tokens: Organizations that operated an affected version (v0.0.12 to v0.0.18) must assume exposure and immediately revoke and rotate any InfluxDB API tokens that were configured for the server. According to the advisory, there is no workaround that fully fixes the issue without upgrading.
Until upgrades are completed, users should restrict the server from operating with MCP clients that process untrusted external content and ensure tokens adhere to least-privilege principles.
Implications for AI-Integrated Backend Security
This vulnerability highlights the security gaps that can emerge when pairing AI assistants with third-party backend tooling. Even without direct user malicious intent, hidden instructions processed inside web documents or chat inputs can manipulate backend infrastructure into leaking sensitive credentials.
Coming on the heels of other recent AWS Labs MCP advisories—such as the path traversal flaw in the AWS Pricing MCP Server and policy bypasses in the AWS API MCP Server—this incident serves as a crucial reminder for organizations deploying AI-driven developer workflows to rigorously audit their credential handling and endpoint validation posture.
Sources
- AWS Security Bulletin: 2026-185-AWS
- GitHub Security Advisory: GHSA-2f9r-g4f3-wc2j (awslabs/mcp)
- NVD Entry: CVE-2026-126681
- PyPI Release: awslabs.timestream-for-influxdb-mcp-server 0.0.19