/

CVE-2026-55534

Severity:

HIGH RISK

(8.6)

Missing Authentication (CWE-306) in PraisonAI Agents

CVE-2026-55534

Deferred

Vulnerability type

Category:

Missing Authentication

CWE:

CWE-306

Impact:

Unauthorized agent invocation and access to agent-mediated operations

Severity:

HIGH RISK

(8.6)

CVSS version:

3.1

Catch this CVE while code is being written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Summary

PraisonAI's agent-serving API can accept requests to invoke configured agents without verifying credentials, even when an API key is supplied to the server. A network caller who can reach the affected endpoints may trigger agent actions and access results without the intended authentication control.

Affected context

The issue occurs in PraisonAI deployments using the praisonai serve agents command during the affected release range, where an API key is accepted as a startup option but authentication is not enforced for the agent POST endpoints. Risk depends on network exposure, the agents configured by the deployment, and the permissions or data available to those agents.

Exploit preconditions

  • Attack surface: Network-accessible PraisonAI agent-serving HTTP API
  • Required access: Network connectivity to the service; a valid API key is not required by the affected endpoints
  • Trigger point: POST requests to /agents or /agents/{agent_name}
  • Why exploitation is possible: The endpoint handlers do not authenticate callers before processing agent invocation requests

Why this happens

The server accepts an API-key configuration value without consistently applying an authentication check to the agent POST routes. This creates a mismatch between the documented or configured security boundary and the behavior of the request handlers. The pattern maps to CWE-306, Missing Authentication for Critical Function, because sensitive functionality remains reachable without verifying the identity of the caller.

Potential impact

  • Unauthorized callers may invoke configured agents
  • Agent outputs or connected data may be exposed
  • Agents may perform unintended actions using their configured permissions
  • Costs, resource consumption, or downstream service activity may increase
  • An attacker could potentially use agent capabilities to affect connected systems

General mitigation guidance

Mitigation generally focuses on enforcing authentication at every route that

performs or initiates privileged agent operations.

Common defensive patterns include:

  • Applying centralized authentication middleware to all agent invocation endpoints
  • Requiring valid credentials before request parsing or agent execution
  • Returning consistent authorization failures for unauthenticated requests
  • Separating authentication from optional API-key configuration so security controls cannot be silently bypassed
  • Restricting network exposure through private interfaces, gateways, or authenticated reverse proxies
  • Running agents with least-privilege identities and narrowly scoped access to tools and data
  • Limiting agent capabilities and applying authorization checks to sensitive operations

How teams detect issues like this

Organizations can look for:

  • Static analysis and code review checks confirming authentication middleware covers every agent POST route
  • Tests that submit requests without credentials and verify that execution does not occur
  • Route inventories compared against authentication and authorization policies
  • API gateway or service-mesh policies requiring identity verification for agent endpoints
  • Monitoring for unauthenticated requests, unusual invocation volume, and agent activity without corresponding identity records
  • Automated security review tools that flag configured authentication controls which are not enforced by request handlers

Sources

SourceLink
github.comgithub.com/MervinPraison/PraisonAI/commit/2f9677abb2ea68eab864ee8b6a828fd0141612e1
github.comgithub.com/MervinPraison/PraisonAI/releases/tag/v4.6.58
github.comgithub.com/MervinPraison/PraisonAI/security/advisories/GHSA-7ww9-85pg-cv4x

Frequently asked

questions

Yes, when the PraisonAI agent service is reachable by an attacker over the

network. Exploitation does not require credentials at the affected endpoints.

No. The issue is that the API-key option can be supplied while POST requests

to the agent endpoints still proceed without authentication.

Require authentication and, where appropriate, authorization in shared

middleware that covers every route capable of invoking an agent before the

request reaches agent execution logic.

Maintain route-level authentication tests, review configuration-to-enforcement

mappings, minimize network exposure, and run agents with least privilege.

Find out what's already
exploitable in your codebase.