---
id: "CVE-2026-109437"
kind: "vulnerability"
aliases: ["CVE-2026-109437"]
title: "Cognition: unauthenticated API lets remote users run shell commands"
severity: "critical"
cvss: 9.8
epss: null
exploited: false
status: "open"
category: "ai-tooling"
products: [{"name": "CognicellAI/Cognition", "fixed": ["0.13.0"], "affected": "< 0.13.0", "ecosystem": "github"}]
ecosystems: ["github"]
tags: ["ai-agents", "ai-tooling", "cognition", "deep-agents", "mcp-adjacent", "self-hosted"]
not_affected_if: ["api-not-exposed", "version-0.13.0-or-later"]
published_at: "2026-10-19T19:17:19+00:00"
first_seen_at: "2026-10-23T21:57:18+00:00"
updated_at: "2026-10-23T22:39:56+00:00"
sources: [{"source": "nvd", "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-109437", "first_seen_at": "2026-10-23T21:57:18+00:00"}, {"source": "analyst", "url": null, "first_seen_at": "2026-10-23T22:28:34+00:00"}]
---

# Cognition: unauthenticated API lets remote users run shell commands

Cognition before 0.13.0 ships its REST/SSE API with no authentication and defaults its agent sandbox to local execution. An attacker who can reach the API can create a session, send a message, and have the agent run shell commands as the server process.

_Summary, check and fix written by a model from the sources; labelled as such._

## Affected

- github:CognicellAI/Cognition affected < 0.13.0 fixed 0.13.0

## How to check

Check the installed Cognition version; versions below 0.13.0 are affected. Confirm whether the API port is reachable from untrusted networks.

## Fix

Upgrade Cognition to 0.13.0 or later. Until then, block external access to the API and do not run the local shell backend on a shared host.

## Timeline

- 2026-10-23T21:57:18+00:00 nvd https://nvd.nist.gov/vuln/detail/CVE-2026-109437
- 2026-10-23T22:28:34+00:00 analyst — Lasso Security advisory (researchers' own write-up, with PoC steps) and VulnCheck CNA record read; fix v0.13.0 per both. No public chatter on X; exploited in the wild not reported (KEV no). Lasso's PoC is a local harness; I did not retrieve the repo release page, so the fix tag was not read directly.

## Documents

- [advisory] Lasso Security: CVE-2026-109437 Unauthenticated RCE via Local Shell Backend https://www.lasso.security/vulnerability-database/cve-2026-109437
- [advisory] VulnCheck: Cognition < 0.13.0 Unauthenticated RCE https://www.vulncheck.com/advisories/cognition-unauthenticated-rce-via-local-shell-backend
- [mentions] Lasso Security: five agent platforms ship an ungated shell tool https://theinjection.dev/items/lasso-agent-platform-shell-tool-cves/

## From the sources

Cognition before 0.13.0 contains an unauthenticated remote code execution vulnerability due to defaulting its agent sandbox backend to local execution and shipping no authentication on its REST/SSE API. Attackers can create a session via POST /sessions and post a message via POST /sessions/{id}/messages without credentials, causing the Deep Agents ReAct loop to run attacker-directed shell commands through the local shell backend, resulting in remote code execution as the server process.

---

_Free to read at a human pace. An API with keys and an MCP server are in preparation; early access: hello@infrastack.ai_