Skip to main content

mcp-memory-service EUVDEUVD-2026-38058

| CVE-2026-49291 HIGH
Missing Authorization (CWE-862)
2026-06-19 GitHub_M GHSA-2r68-g678-7qr3
8.1
CVSS 3.1 · Vendor: GitHub_M
View

Attack ChainAIDerived

Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain

  1. Access
    Obtain read-scope OAuth token
  2. Delivery
    Reach /mcp JSON-RPC endpoint over network
  3. Exploit
    Send tools/call for delete_memory or store_memory
  4. Execution
    Scope check passes on read only
  5. Persist
    Handler executes mutating operation
  6. Impact
    Memory store tampered or wiped

Vulnerability Assessment

Weakness
CWE
OWASP Top 10:2025
A01:2025 - Broken Access Control
ASVS 5.0
V8 - Authorization Authorization checks on every request
MITRE ATT&CK

1 technique across 1 tactic: Privilege Escalation

The technique list with ATT&CK links and the kill-chain view is part of the Basic plan and up.

Exploitation Conditions

Requires a valid OAuth access token bearing the read scope for the target mcp-memory-service instance and network reachability to the HTTP /mcp JSON-RPC endpoint; the instance must be running a vulnerable version prior to 10.65.3 with the MCP HTTP interface enabled (it is the affected code path, distinct from the REST API). No user interaction and no write/admin scope are needed, and exploitation works against default configurations of the HTTP MCP endpoint. Limiting factors: deployments that never issue read-only tokens, that disable the HTTP MCP endpoint, or that front it with an external authorization proxy are not exposed.

Impact

The CVSS 3.1 vector AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:H/A:H (8.1 High) accurately reflects a network-reachable, low-complexity bypass that requires a low-privileged OAuth token and yields high integrity and availability impact with no confidentiality loss - consistent with an attacker overwriting or deleting memories but not reading new data they could not already access. There is no CISA KEV entry, no public POC referenced, and no EPSS figure supplied, so real-world exploitation pressure appears low at present; however the prerequisite is modest (any issued read-scope token), so any deployment that hands out read tokens to semi-trusted agents or integrations is materially exposed. Defenders should treat this as a real priority for any multi-tenant or agent-shared deployment, and a lower priority for single-user local installs where only the operator holds tokens.

Exploit ScenarioDerived

An attacker who has obtained or been legitimately issued an OAuth token with only the read scope - for example a third-party AI agent integration meant to query memories - sends a JSON-RPC tools/call request to /mcp invoking delete_memory or store_memory with attacker-chosen content. Because the endpoint only verifies the read scope, the call succeeds and the attacker can wipe stored memories or poison them with adversarial content that downstream LLM consumers will later retrieve. No public POC is referenced, but the bypass is trivial to reproduce given the description.

This Week Schedule remediation this week. Why: HIGH severity (CVSS 7.0-8.9).
Patch available 0.3% exploitation probability (EPSS)

Get an e-mail when this CVE is exploited, patched or re-scored. Free account.

Alert me

AnalysisAIv1 – Jun 19, 2026

Authorization bypass in doobidoo mcp-memory-service prior to 10.65.3 lets OAuth clients with only the read scope invoke mutating MCP tools such as store_memory and delete_memory via the /mcp JSON-RPC endpoint, despite the equivalent REST endpoints correctly enforcing the write scope. Authenticated read-only clients can therefore tamper with or destroy memory entries used by downstream AI applications. No public exploit identified at time of analysis, and the issue is not listed in CISA KEV.

RemediationAI

Vendor-released patch: upgrade mcp-memory-service to 10.65.3 or later via pip (pip install --upgrade mcp-memory-service>=10.65.3), as confirmed by GHSA-2r68-g678-7qr3 and the PyPI release at https://pypi.org/project/mcp-memory-service/10.65.3. If immediate patching is not possible, compensating controls include disabling or network-blocking the /mcp JSON-RPC endpoint while continuing to use the REST API which already enforces write scope correctly, or issuing OAuth tokens only with the write scope to trusted clients and revoking any outstanding read-only tokens issued to untrusted agents (side effect: read-only integrations lose MCP access entirely until upgrade). Placing the service behind a reverse proxy that rejects MCP tools/call requests for store_memory and delete_memory is a finer-grained stopgap but requires JSON-RPC body inspection and will break legitimate write clients that use MCP.

Affected VersionsENISA

mcp-memory-service < 10.65.3

Affected Packages

pypi mcp-memory-service fix: 10.65.3

Source: OSV (Open Source Vulnerabilities)

Affects your stack

Affects 1 package in pypi

Checking this CVE against your own stacks and servers is part of the Basic plan and up.

Technical ContextAI · show details

mcp-memory-service is a Python package (CPE cpe:2.3:a:doobidoo:mcp-memory-service) that provides a semantic memory layer for AI/LLM applications and exposes both a REST API and an MCP (Model Context Protocol) JSON-RPC interface. The MCP endpoint at /mcp gates every request behind a single OAuth scope check requiring only 'read', then dispatches the tools/call method directly to per-tool handlers without re-evaluating the scope appropriate to each tool. This is a textbook CWE-862 Missing Authorization defect: the scope-to-operation mapping enforced on REST routes was never mirrored on the MCP dispatcher, so write-class tools like store_memory and delete_memory inherit the relaxed read gate.

Know when Mcp Memory Service gets its next vulnerability

  • Watch products and get an email when something new is published
  • Check this CVE against your own stack and servers
  • Exploit and PoC references, threat intelligence from 22 sources

No credit card · 7-day full trial

Severity by source

Vendor (GitHub_M) PRIMARY 8.1 HIGH
AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:H/A:H
vuln.today AI 8.1 HIGH

Network-reachable HTTP endpoint, trivial JSON-RPC call (AC:L), requires a low-privileged read-scope OAuth token (PR:L), no user interaction; impact is write/delete of memories so I:H and A:H, no new confidentiality disclosure (C:N).

3.1 AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:H/A:H
4.0 AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:H/VA:H/SC:N/SI:N/SA:N

Primary rating from Vendor (GitHub_M).

CVSS VectorVendor: GitHub_M

Attack Vector
Network
Attack Complexity
Low
Privileges Required
Low
User Interaction
None
Scope
Unchanged
Confidentiality
None
Integrity
High
Availability
High

Lifecycle Timeline

1
Patch available
Jun 19, 2026 - 21:02 EUVD

DescriptionCVE.org

mcp-memory-service is a semantic memory layer for AI applications. Prior to version 10.65.3, the HTTP MCP JSON-RPC endpoint at /mcp requires only OAuth read scope for all requests, then dispatches tools/call directly to handlers that include mutating tools. A read-only OAuth client can call store_memory and delete_memory through MCP even though the corresponding REST endpoints require write scope. Version 10.65.3 patches the issue.

Executive ViewAI

Why It Matters

doobidoo mcp-memory-service contains an authorization flaw allowing read-only API credentials to invoke data-modifying operations, jeopardizing the integrity of memory systems that downstream AI applications depend on. This affects any organization using the service for AI application memory management, creating immediate exposure to unauthorized data tampering and destruction. No vendor patch is currently available to resolve the vulnerability.

Business Impact: High
Compensating Controls

No vendor-released patch identified at time of analysis. Interim mitigations: (1) Enforce write-scope validation at API gateway or proxy layer for all /mcp JSON-RPC mutating operations; (2) Restrict read-only OAuth token credentials to read-only REST endpoints only, blocking /mcp access; (3) Enable comprehensive audit logging on all memory mutation operations to detect unauthorized access; (4) Network-segment memory service from internet-facing and untrusted networks where feasible; (5) Immediately rotate all read-scope OAuth credentials assigned to external or integrated applications.

Threat Intelligence Overview

EPSS Score
0.3%
Top 82% exploitable
Threat Intel
1
1 sources
Exploit
POC
Public exploit code exists
Data Quality
6 sources
Confidence: High
EUVD NVD CVE.org EPSS Vulnrichment GHSA GitHub

How this decision was reached

The engine reads each signal below and applies a fixed list of rules from top to bottom. The first rule that matches sets the recommended action – like a statement, every line is shown.

SignalValue
Severity HIGH
CISA KEV No
Public POC Yes
EPSS 0.3%
Patch available Yes
Active threat No
Deciding rule
HIGH severity (CVSS 7.0-8.9)
What every action level means
Emergency Immediate action required. Criteria: CRITICAL + in CISA KEV (actively exploited) without patch, OR CRITICAL + EPSS > 30% without patch, OR CRITICAL + active threat signals without patch.
Act Now Prioritize within 24-48h. Criteria: in KEV with patch available, OR CRITICAL + public POC, OR HIGH + EPSS > 10%, OR CRITICAL + active threat, OR any CRITICAL severity.
This Week Schedule remediation this week. Criteria: HIGH severity (CVSS 7.0-8.9). Patch within normal change window.
This Month Plan within 30 days. Criteria: MEDIUM severity (CVSS 4.0-6.9), lower exploitation probability. Include in regular patch cycle.
Monitor Track and reassess. Criteria: LOW severity (CVSS 0.1-3.9), minimal exploitation risk. No immediate action needed.

SSVC AssessmentENISA

SSVC Decision
Scheduled
Exploitation
None
System Exposure· derived
Open
Automatable
No
Human Impact
High
Technical Impact
Total

Priority Score

60
Low Medium High Critical
KEV: 0
EPSS: 0
CVSS: +40
PoC: +20
Weaponized: 0
Threat intel: 0
Gov advisory: 0

Attack Technique

Authentication bypass attacks exploit flaws in the verification mechanisms that control access to systems and applications.

This site uses cookies essential for authentication and security. No tracking or analytics cookies are used. Privacy Policy