# CVE-2026-115727: AgenticSeek Flask debug mode remote code execution

Source: https://junglewise.ai/threats/cve-2026-115727-agenticseek-before-commit-409840c-contains-a-remote-code  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-10-23. Updated: 2026-10-23  
License: CC BY 4.0 with a link back

CVE-2026-115727 · Severity: high · CVSS 8.1 · Published 2026-10-23

AgenticSeek is an AI-powered search application that uses a Flask web server to handle language model requests. The vulnerability allows attackers on the network to execute arbitrary Python and operating system commands by exploiting hardcoded debug mode and deriving the Werkzeug debugger PIN, potentially leading to complete server compromise and data exfiltration.

Technologies: [Fosowl Agenticseek](https://junglewise.ai/threats/technologies/agenticseek). Vendors: [Fosowl](https://junglewise.ai/threats/vendors/fosowl).

## Quick answers

**Is CVE-2026-115727 exploited in the wild?** No. Junglewise Threat Intelligence has no report of CVE-2026-115727 being exploited in the wild.

**How severe is CVE-2026-115727?** High severity, with a CVSS score of 8.1, published 23 October 2026.

**Which products and versions are affected?** Fosowl Agenticseek before commit 409840c.

**Is there a fix for CVE-2026-115727?** Yes: fixed in AgenticSeek commit 409840c and later. Until then: Disable Flask debug mode by setting debug=False in app.run() calls, and restrict network access to the llm_server port to trusted hosts only using a firewall or network segmentation. Junglewise finds every place you run it, drafts the upgrade and applies it once you approve.

**How is CVE-2026-115727 attacked?** It is attacked over the network, in a software library.

**Am I affected by CVE-2026-115727?** Check whether you run Fosowl Agenticseek at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

## How to fix

- Fixed in AgenticSeek commit 409840c and later

Disable Flask debug mode by setting debug=False in app.run() calls, and restrict network access to the llm_server port to trusted hosts only using a firewall or network segmentation.

## Technical details

The llm_server/app.py file hardcodes Flask debug mode (debug=True) while binding to all network interfaces (0.0.0.0), enabling the Werkzeug interactive debugger. An attacker with network access to the server port can derive the debugger PIN and gain remote code execution with llm_server process privileges. The fix disables debug mode in production deployments.

## Affected products

- Fosowl AgenticSeek before commit 409840c

## Timeline

- 2026-10-23: disclosed

## References

- [https://github.com/Fosowl/agenticSeek/blob/d238f8e5de10e4ece40d93d80fb4daa4c8a2f34b/llm_server/app.py](https://github.com/Fosowl/agenticSeek/blob/d238f8e5de10e4ece40d93d80fb4daa4c8a2f34b/llm_server/app.py)
- [https://github.com/Fosowl/agenticSeek/commit/409840c033851d861c5c83a91c05fd4a2a75a096](https://github.com/Fosowl/agenticSeek/commit/409840c033851d861c5c83a91c05fd4a2a75a096)

## Related threats

- [AgenticSeek unauthenticated remote code execution in query API](https://junglewise.ai/threats/cve-2026-72776-agenticseek-unauthenticated-remote-code-execution-in-query-api)
- [Fosowl agenticSeek code injection in PyInterpreter.execute](https://junglewise.ai/threats/cve-2026-5584-fosowl-agenticseek-code-injection-in-pyinterpreter-execute)