Skip to main content
Vulnerability Database/CVE-2026-77937

CVE-2026-77937: Letta Auth Bypass Vulnerability

CVE-2026-77937 is an authentication bypass flaw in Letta that lets attackers impersonate users via HTTP headers to access agents and memory. This post explains its impact, affected versions, and mitigation steps.

Published:

Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.

Sources: National Vulnerability Database, CISA KEV, Exploit-DB

CVE-2026-77937 Overview

CVE-2026-77937 is an authentication bypass vulnerability in Letta through version 0.16.8. The flaw allows API callers to impersonate other users by supplying a self-asserted user_id HTTP header. Attackers can set user_id to another tenant's identifier, even with LETTA_SERVER_SECURE enabled, to read and modify that tenant's agents, memory, and configuration. The weakness is classified as [CWE-639] Authorization Bypass Through User-Controlled Key. The vulnerability affects multi-tenant deployments of the Letta agent framework and enables cross-tenant data access over the network with low-privileged credentials.

Critical Impact

Any authenticated API caller can impersonate arbitrary tenants and gain full read and write access to other users' agents, memory, and configuration.

Affected Products

  • Letta versions through 0.16.8
  • Letta REST API server component (letta/server/rest_api/dependencies.py)
  • Multi-tenant Letta deployments including those with LETTA_SERVER_SECURE enabled

Discovery Timeline

  • 2026-10-20 - CVE-2026-77937 published to NVD
  • 2026-10-20 - Last updated in NVD database

Technical Details for CVE-2026-77937

Vulnerability Analysis

Letta is an open-source agent framework that exposes a REST API for managing agents, memory, and configuration on a per-tenant basis. The server resolves the identity of the calling user from an HTTP request header named user_id. The dependency responsible for user resolution, defined in letta/server/rest_api/dependencies.py, trusts the client-supplied header without validating that the authenticated principal is authorized to act as that user.

As a result, any caller who can reach the API, including a caller with low-privileged credentials when LETTA_SERVER_SECURE is enabled, can set user_id to the identifier of a different tenant. The server then processes the request under the impersonated identity. This grants cross-tenant read and write access to agents, long-term memory, and runtime configuration belonging to other users.

Root Cause

The root cause is authorization through a user-controlled key [CWE-639]. The server relies on a self-asserted identifier transmitted in the user_id header as the sole basis for scoping requests to a tenant. The authentication layer gated by LETTA_SERVER_SECURE validates that a caller holds a valid credential but does not bind that credential to a specific tenant. The missing binding between the authenticated session and the asserted user_id allows horizontal privilege escalation across tenants.

Attack Vector

An attacker needs network reach to the Letta API and, where enforced, any valid API credential accepted by LETTA_SERVER_SECURE. The attacker issues standard REST requests to agent, memory, or configuration endpoints and sets the user_id header to the target tenant's identifier. The server returns and mutates data belonging to that tenant. No user interaction is required, and the attack can be scripted against enumerated tenant IDs.

Readers can review the vulnerable dependency in the Letta dependencies.py source, the Lasso advisory for CVE-2026-77937, and the VulnCheck advisory for Letta.

Detect Identity Attacks the Moment Authentication Fails

Singularity™ Identity surfaces credential misuse, lateral movement, and privilege escalation in real time, so you can contain the breach before it spreads.

Stop Identity Attacks

Detection Methods for CVE-2026-77937

Indicators of Compromise

  • API requests containing an explicit user_id HTTP header where the asserted identifier does not match the authenticated principal.
  • Access to agent, memory, or configuration endpoints across multiple distinct user_id values originating from a single source IP, API key, or session.
  • Unexpected modifications to agent state, memory records, or configuration owned by tenants that have not recently logged in.

Detection Strategies

  • Log every incoming user_id header value alongside the authenticated principal and alert when the two diverge.
  • Baseline normal tenant-to-caller mappings and flag deviations where one caller touches multiple tenant IDs within a short window.
  • Inspect reverse proxy and application access logs for enumeration patterns against /agents, /memory, and configuration endpoints.

Monitoring Recommendations

  • Forward Letta API access logs and application logs to a centralized analytics pipeline for correlation with authentication events.
  • Monitor outbound data volumes from Letta API endpoints for anomalous spikes consistent with bulk cross-tenant extraction.
  • Track administrative and configuration-change events per tenant and alert on changes initiated by unexpected principals.

How to Mitigate CVE-2026-77937

Immediate Actions Required

  • Restrict network access to the Letta API to trusted clients using firewall rules, private networking, or an authenticating reverse proxy.
  • Strip or reject the user_id HTTP header at an upstream proxy so that it cannot be supplied by API clients.
  • Rotate any API credentials that may have been exposed and review audit logs for cross-tenant access attempts.
  • Treat multi-tenant Letta deployments as compromised until logs confirm no impersonation occurred.

Patch Information

No fixed version is listed in the NVD record at the time of publication. The advisory identifies all releases through 0.16.8 as affected. Operators should monitor the VulnCheck advisory for Letta and the Letta repository for a patched release and upgrade as soon as a corrected version is available.

Workarounds

  • Deploy a reverse proxy such as NGINX or Envoy in front of Letta that unconditionally removes the user_id request header before forwarding.
  • Operate a dedicated Letta instance per tenant to eliminate the cross-tenant impersonation surface entirely.
  • Enforce mutual TLS or network-level segmentation so that only trusted internal services can reach the Letta API.
  • Enable verbose request logging and alerting on any request containing a user_id header until a vendor patch is applied.
bash
# Example NGINX configuration to strip the user_id header before proxying to Letta
server {
    listen 443 ssl;
    server_name letta.internal.example.com;

    location / {
        proxy_set_header user_id "";
        proxy_pass http://letta_backend:8283;
    }
}

Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.

Sources: National Vulnerability Database, CISA KEV, Exploit-DB

Experience the Most Advanced Cybersecurity Platform

See how the world’s most intelligent, autonomous cybersecurity platform can protect your organization today and into the future.