HOL LogoGuard

Explore HOL

  • HOL home
  • AI agent registry
  • AI plugins
  • Open standards
  • HOL members
  • Sponsor HOL
  • Advertise with HOL

Guard product

  • Guard overviewLocal security and control for AI agents and the tools they use.
  • FeaturesRuntime protection, policy routing, review, and evidence.
  • CIGARPoint your coding agent at the right lines and cut its token bill by about a third.

Explore Guard

  • Product previewWalk through Guard surfaces in read-only demo mode.
  • ComparisonCompare Guard with native controls and AI security vendors.

AI tools

  • All AI toolsEvery supported AI tool and how Guard applies policy to it.
  • Codex
  • Claude Code
  • Cursor
  • Antigravity CLI
  • OpenCode
  • Hermes
  • OpenClaw
  • GitHub Copilot CLI
  • Antigravity
  • Kimi
  • Grok
  • Pi / Oh My Pi
  • Zcode

Extensions

  • All extensionsBrowse command and MCP coverage with owners and stated limits.
  • Command coverageShell command protection across clouds, databases, backups, and packages.
  • MCP server coverageSee how Guard maps risk state across MCP tools and servers.
  • Core safetyThe safety floor listings that ship with Guard.
  • Data and resilienceBackup and storage command protection.
  • Cloud and infrastructureAWS, Azure, GCP, Kubernetes, and more.

Security

  • AI security hubSecurity research, advisories, and agent safety coverage.
  • AI tool securitySecurity profiles for each supported coding agent.
  • Safe labsHands-on attack simulations with safe boundaries.
  • Redacted warningsReal blocked actions with sensitive details removed.
  • AdvisoriesCoordinated disclosure reports for AI tooling.
  • Active CVEsSearch active CVEs affecting AI tooling.

Learn

  • Security guidesPractical guides for securing AI agent workflows.
  • DocsInstall, configure, and operate Guard with confidence.
  • ResearchPublished security research, benchmarks, and methodology.
  • Gauntlet benchmarkGuard tested with real coding agents: safe work runs, risky actions stop.

Community

  • ReleasesVersion history, shipped changes and upgrade notes.
  • ContributorsThe people and contributions behind HOL Guard.
  • AffiliatesShare Guard with your audience and earn from referrals.
  • SponsorFund agreed research, implementation and shared testing through the HOL Partner Program.
  • AdvertiseCompare README logo placements, previews and monthly prices, or promote a plugin or extension.
PricingEnterpriseOpen AppInstall Guard
  1. Guard
  2. Security
  3. CVEs
  4. CVE 2026 49257 mcp pinot unauthenticated tool invocation via
HOL Guard

Public security guidance for teams protecting AI harnesses, MCP servers, skills, prompts, and local tool execution.

Install Guard

AI Security

  • Prompt injection
  • MCP security
  • OWASP MCP mapping
  • Supply chain

Resources

  • Trust packet
  • Harness setup
  • Redacted warnings
  • Safe labs

Product

  • Install Guard
  • Pricing
  • Open dashboard

Get HOL Guard for your computer

HOL Guard is a free desktop app that watches what coding agents such as Claude Code, Codex and Cursor run, read and install, and asks you before anything risky happens.

  • Download for macOS
  • Windows download
  • Linux download

Free. macOS 13+ on Apple silicon, Windows x64, Linux AppImage. Other ways to install

Star HOL Guard on GitHub to download
Runs on
macOS 13 or later on Apple silicon, Windows x64 and Linux x86_64.
Price
Free. Paid plans add Guard Cloud for teams. See pricing
Where your data stays
On your computer. Guard Cloud is optional.
Token savings
58.6% to 61.5% fewer tokens across three coding agents in our CIGAR benchmark.

HOL

HOL builds open infrastructure for AI agents: HOL Guard, a directory of AI plugins and MCP servers, an AI agent registry and the open HCS standards.

  • GitHub (opens in a new tab)
  • X (Twitter) (opens in a new tab)
  • YouTube (opens in a new tab)
  • Telegram (opens in a new tab)

HOL Guard

  • What is HOL Guard
  • Features
  • Cut coding agent token costs (CIGAR)
  • Install Guard
  • Download the desktop app
  • Pricing
  • Guard for teams
  • Compare Guard
  • Release notes

Guard for your AI tool

  • Guard for Claude Code
  • Guard for Codex
  • Guard for Cursor
  • Guard for GitHub Copilot CLI
  • Guard for OpenCode
  • Guard for Pi and Oh My Pi
  • Guard for Hermes
  • Guard for OpenClaw
  • All supported AI tools

Research and security

  • Gauntlet: Guard tested with real agents
  • CIGAR token benchmark
  • Research
  • AI security hub
  • Security advisories
  • Active CVEs in AI tools
  • AI agent security guides
  • Extensions and coverage

Plugins and MCP servers

  • Browse AI plugins
  • New plugin launches
  • Best Claude Plugins
  • Best Codex Plugins
  • Best Grok Plugins
  • Best Kimi Plugins
  • Best DeepSeek Plugins
  • Best Antigravity Plugins
  • Best MCP Servers
  • Best Cursor Plugins
  • Best OpenCode Plugins

AI agent registry

  • Best AI agents
  • Best ERC-8004 agents
  • Best Virtuals agents
  • Best A2A agents
  • Best x402 payable agents
  • Submit an ERC-8004 contract
  • Feature your agent

Developers

  • Guard docs
  • Documentation index
  • Developer hub
  • Registry API reference
  • OpenAPI spec
  • Registry OpenAPI spec
  • Run in Postman (opens in a new tab)
  • HCS standards

Company

  • About HOL
  • Blog
  • Contribute to HOL
  • Guard contributors
  • Sponsor
  • Advertise
  • Affiliate program
  • Contact

Copyright ยฉ 2026 HOL DAO LLC. All rights reserved.

  • Privacy
  • Terms of Service
Back to active CVEs
Critical ยท CVSS 10.0CVE-2026-49257GHSA-73CV-556C-W3G6

mcp-pinot: Unauthenticated tool invocation via default oauth_enabled=False + host 0.0.0.0 bindCVE-2026-49257

Answer in brief

CVE-2026-49257 records a Critical severity (CVSS 10.0) vulnerability in mcp-pinot: Unauthenticated tool invocation via default oauth_enabled=False + host 0.0.0.0 bind. The current sources do not mark it as known exploited. The current feed maps mcp-pinot-server (pip), mcp-pinot-server (pypi). Check affected ranges and fixed versions before updating.

Analysis pending evidence review

HOL Guard separates source facts from reviewed analysis. See the methodology.

Published Jun 18, 2026Updated Nov 16, 2026Source checked Nov 16, 2026First seen by HOL Jun 22, 2026Material review Nov 16, 2026
Upstream Advisory

Key facts

Risk
Critical ยท CVSS 10.0
Exploitation
Not marked as known exploited
Affected software
2 mapped packages or products
Fix availability
Available

Why this deserves its current priority

CVSS is 10.0. The current sources do not mark it as known exploited. Treat this as a source-backed prioritization signal, not a statement about your environment.

Analysis status

Analysis pending evidence review

Factual feed record only; HOL analysis is not approved for indexing. Read the methodology.

Affected scope and exposure questions

The current feed maps mcp-pinot-server (pip), mcp-pinot-server (pypi). Check affected ranges and fixed versions before updating.

Mapped affected packages and fixed versions
PackageAffected rangeFixed version
mcp-pinot-serverpip<=3.0.13.1.0
mcp-pinot-serverpypi>=0 <3.1.03.1.0

Recommended response

  1. 1Check inventory. Check lockfiles and deployed manifests for mcp-pinot-server, mcp-pinot-server.
  2. 2Review the reported fix. Update mcp-pinot-server to 3.1.0; mcp-pinot-server to 3.1.0 if you use the affected versions. Test the change in a non-production environment first.

Evidence timeline and material changes

  1. Published upstream

    Jun 18, 2026

    Evidence: source:ghsa:source_dates:source-dates:record
  2. Source modified

    Nov 16, 2026

    Evidence: source:ghsa:source_dates:source-dates:record
  3. First seen by HOL

    Jun 22, 2026

Sources and claim methodology

  • GitHub security advisorygithub.com
  • GitHub security advisorygithub.com
  • GitHub security advisorygithub.com
  • GitHub security advisorygithub.com
  • GitHub security advisorygithub.com
  • GitHub security advisorygithub.com
  • NVD vulnerability recordnvd.nist.gov
  • Source referencepypi.org
  • GitHub security advisorygithub.com
Upstream source description

## Resolution Fixed in [v3.1.0](https://github.com/startreedata/mcp-pinot/releases/tag/v3.1.0), released 2026-05-25. The fix was merged in [PR #95](https://github.com/startreedata/mcp-pinot/pull/95) at commit [`1c7d3f9`](https://github.com/startreedata/mcp-pinot/commit/1c7d3f9cd384854bf72c127d230bdb32299475ad). The fix changes the default HTTP bind host to `127.0.0.1`, refuses non-loopback HTTP/HTTPS exposure unless OAuth is enabled, makes Helm exposure opt-in and OAuth-gated, and adds parser-backed single-statement read-only validation for `read-query`. ## CVSS evaluation Reviewed on 2026-05-25. The advisory remains **Critical** with `CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H` = **10.0**. Rationale: | Metric | Value | Reason | |---|---|---| | AV | Network | The default HTTP server bound to `0.0.0.0:8080` and accepted remote HTTP requests. | | AC | Low | Exploitation required only a direct MCP tool call. | | PR | None | OAuth was disabled by default. | | UI | None | No user interaction was required. | | S | Changed | The vulnerable MCP server used its server-side credentials to act on the separate Pinot cluster security boundary. | | C | High | Unauthenticated callers could read table data and cluster metadata through server-side Pinot credentials. | | I | High | Unauthenticated callers could create or update schemas and table configs where the server-side account had those privileges. | | A | High | Expensive queries and configuration mutations could degrade or disrupt Pinot availability. | # Unauthenticated tool invocation via default oauth_enabled=False + host 0.0.0.0 bind ## Summary `mcp-pinot` v3.0.1 (and earlier) defaults to running an HTTP MCP server bound to `0.0.0.0:8080` with no authentication enabled. All MCP tools, including SQL query execution, schema creation, and table-config mutation, are reachable by any network-adjacent caller. The server proxies these calls using server-side Pinot credentials, producing a confused-deputy condition that yields full read/write access to the configured Pinot cluster. ## Affected versions - All releases on `main`, confirmed in tags v2.1.0 through v3.0.1. - Affected files: `mcp_pinot/server.py`, `mcp_pinot/config.py`. ## Root cause Three defaults compose to produce unauthenticated network exposure: **1. Auth is opt-in and defaults to off** (`mcp_pinot/config.py:64,328`): ```python @dataclass class ServerConfig: ... oauth_enabled: bool = False ... def load_server_config() -> ServerConfig: return ServerConfig( ... oauth_enabled=os.getenv("OAUTH_ENABLED", "false").lower() == "true", ... ) ``` **2. Auth construction is gated by `oauth_enabled`** (`mcp_pinot/server.py:26-46`): ```python _auth = None if server_config.oauth_enabled: oauth_config = load_oauth_config() token_verifier = JWTVerifier(...) _auth = OAuthProxy(...) mcp = FastMCP("Pinot MCP Server", auth=_auth) ``` When `oauth_enabled` is false (default), `_auth` stays `None` and `FastMCP` registers all `@mcp.tool` endpoints with no authentication. **3. Default bind is all interfaces on a well-known port** (`mcp_pinot/config.py:60-61`): ```python host: str = "0.0.0.0" port: int = 8080 ``` The HTTP transport in `server.py:263-268` uses these values directly. Any operator following the README's HTTP transport instructions (`uv pip install`, `.env` from `.env.example`, run) ends up with a network-reachable MCP server with no auth. ## Confused-deputy The Pinot client uses server-side credentials loaded from environment variables (`mcp_pinot/config.py:285-294, 300-315`). When an unauthenticated MCP caller invokes `read_query` or any other tool, the request is executed with the server's `PINOT_TOKEN` or `PINOT_USERNAME`/`PINOT_PASSWORD`, which is typically a privileged service account. The MCP server effectively launders the caller's lack of identity into the server's privileges against the upstream cluster. ## Exposed tools All 14 tools in `mcp_pinot/server.py` are exposed without auth in the default configuration: | Tool | Impact when unauthenticated | |---|---| | `read_query` | Arbitrary SELECT against any table allowed by server-side filter (or all tables if no filter) | | `list_tables` | Enumerate cluster schemas | | `table_details`, `segment_list`, `segment_metadata_details`, `tableconfig_schema_details`, `index_column_details`, `get_schema`, `get_table_config` | Read cluster metadata | | `create_schema`, `update_schema` | Create or mutate Pinot schemas | | `create_table_config`, `update_table_config` | Create or mutate table configurations | | `reload_table_filters` | Reload server filter file; response leaks `previous_filters` and `new_filters` lists | | `test_connection` | Cluster diagnostics including host, port, scheme, database, and auth-mode | ## Reproduction Minimal reproduction against a default-configured `mcp-pinot` v3.0.1 instance running on `http://victim:8080/mcp`: ```bash # 1. Enumerate tables (no Authorization header) curl -X POST http://victim:8080/mcp \ -H 'Content-Type: application/json' \ -d '{ "jsonrpc":"2.0", "method":"tools/call", "params":{"name":"list_tables","arguments":{}}, "id":1 }' # 2. Read arbitrary table contents (server forwards using its own Pinot credentials) curl -X POST http://victim:8080/mcp \ -H 'Content-Type: application/json' \ -d '{ "jsonrpc":"2.0", "method":"tools/call", "params":{ "name":"read_query", "arguments":{"query":"SELECT * FROM <table> LIMIT 100"} }, "id":2 }' # 3. Create a new schema (write privileges) curl -X POST http://victim:8080/mcp \ -H 'Content-Type: application/json' \ -d '{ "jsonrpc":"2.0", "method":"tools/call", "params":{ "name":"create_schema", "arguments":{ "schemaJson":"{\"schemaName\":\"attacker_schema\",\"dimensionFieldSpecs\":[{\"name\":\"id\",\"dataType\":\"STRING\"}]}" } }, "id":3 }' ``` ## Severity (CVSS 3.1) `CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H` = **10.0 Critical** | Metric | Value | Reason | |---|---|---| | AV (Attack Vector) | Network | Server defaults to bind on `0.0.0.0:8080` | | AC (Attack Complexity) | Low | No special conditions, single HTTP request | | PR (Privileges Required) | None | No authentication required in default config | | UI (User Interaction) | None | Direct unauthenticated call | | S (Scope) | Changed | Vulnerable MCP component grants access to a separate Pinot cluster (different security authority) | | C (Confidentiality) | High | Full read of any table data the server-side account can reach | | I (Integrity) | High | Schema and table-config writes via `create_schema`, `update_schema`, `create_table_config`, `update_table_config` | | A (Availability) | High | Heavy queries, malformed configs, or schema overrides can degrade or break the cluster | If the operator restricts the bind address to `127.0.0.1` via `MCP_HOST`, AV drops to `Local` and the score reduces. But this is not the documented default. ## Suggested remediation Two independent hardenings, both recommended: **A. Refuse to start in an insecure default**, in `server.py` `main()`, fail-closed when: - `transport != "stdio"` - `server_config.oauth_enabled` is `False` - `server_config.host` is not a loopback address (e.g. not in `{"127.0.0.1", "::1", "localhost"}`) Sample: ```python def _is_loopback(host: str) -> bool: return host in {"127.0.0.1", "::1", "localhost"} def main(): ... if server_config.transport != "stdio" and not server_config.oauth_enabled and not _is_loopback(server_config.host): raise SystemExit( "Refusing to start: HTTP transport bound to non-loopback host " f"({server_config.host}) without OAuth. Set OAUTH_ENABLED=true or " "set MCP_HOST=127.0.0.1 for local-only access." ) ... ``` **B. Default `oauth_enabled` to `True`** and require explicit opt-out for local development. This matches the principle of secure-by-default for network-facing services. **C. Document the threat model** in README under a "Production deployment" section, including: - Explicit warning that the server should not be exposed to untrusted networks without OAuth - Recommendation to set `MCP_HOST=127.0.0.1` for stdio/local-only deployments ## Resources - `mcp_pinot/server.py` lines 26-46, 248-269 - `mcp_pinot/config.py` lines 56-65, 318-330 - FastMCP `auth` parameter behavior when `None`: https://github.com/jlowin/fastmcp - The Register, May 13 2026: MCP database flaws across Doris, Pinot, RDS ## Reporter Independent security researcher. Disclosed via GitHub Security Advisory, 2026-05-23.

Quoted source text, attributed separately from HOL analysis.

Related CVEs

  • RedAmon before 7.8.0 Unauthenticated Root Shell via Terminal PTY WebSocketAlso CWE-306
  • Prime CMS through 0.4.0-beta.0 Authentication Bypass via createUser GraphQL MutationAlso CWE-306
  • Prime CMS through 0.4.0-beta.0 Missing Authentication via GraphQL MutationsAlso CWE-306

Record context

Vulnerability class
Missing Auth
EPSS
Not reported
CWE IDs
CWE-306
Source
GitHub Security Advisories
Source checked
Nov 16, 2026
References
9 linked sources
Open source record