Every CVE Search/ Get API key

CVE-2026-58196

ToolHive: SSRF in remote MCP server authentication discovery (host-side, bypasses container isolation)

4.7 MediumFix available

ToolHive is a utility designed to simplify the deployment and management of Model Context Protocol (MCP) servers. Prior to 0.31.0, remote.Handler.Authenticate in pkg/auth/remote/handler.go invokes discovery.DetectAuthenticationFromServer in pkg/auth/discovery/discovery.go, whose host-side HTTP clients trust remote-server-controlled authentication discovery destinations, follow redirects without host or scheme restrictions, and do not consistently block private addresses. A malicious or compromised remote MCP server can place a resource_metadata URL in WWW-Authenticate for ParseWWWAuthenticate to extract, after which FetchResourceMetadata, OIDC issuer discovery, and well-known discovery can issue GET requests to link-local, RFC1918, or other internal services outside the server's container. The user connects to a server that the user intends to use, but the server controls the internal destination; this path does not depend on the separate NAT64 guard omission because the affected clients did not invoke IsPrivateIP, and it differs from the DCR resolver path that already refused redirects. The demonstrated primitive reaches internal-only HTTP services and reachability or error oracles, and can retrieve AWS metadata credentials where IMDSv1 accepts an unauthenticated GET, while IMDSv2 and GCP metadata prerequisites are not satisfied by the demonstrated request. This vulnerability is fixed in 0.31.0.

Attack profile

How easy it is to exploit

Attack vector
Network
Complexity
Low
Privileges needed
None
User interaction
Required

What an attacker can affect

Confidentiality
Low
Integrity
None
Availability
None

Exploit likelihood

0.43%Up 0.1 points since Sep 16, 2026
90%50%10%1%

What each source reports

CVE List Score used 4.7

ToolHive: SSRF in remote MCP server authentication discovery (host-side, bypasses container isolation)

ToolHive is a utility designed to simplify the deployment and management of Model Context Protocol (MCP) servers. Prior to 0.31.0, remote.Handler.Authenticate in pkg/auth/remote/handler.go invokes discovery.DetectAuthenticationFromServer in pkg/auth/discovery/discovery.go, whose host-side HTTP clients trust remote-server-controlled authentication discovery destinations, follow redirects without host or scheme restrictions, and do not consistently block private addresses. A malicious or compromised remote MCP server can place a resource_metadata URL in WWW-Authenticate for ParseWWWAuthenticate to extract, after which FetchResourceMetadata, OIDC issuer discovery, and well-known discovery can issue GET requests to link-local, RFC1918, or other internal services outside the server's container. The user connects to a server that the user intends to use, but the server controls the internal destination; this path does not depend on the separate NAT64 guard omission because the affected clients did not invoke IsPrivateIP, and it differs from the DCR resolver path that already refused redirects. The demonstrated primitive reaches internal-only HTTP services and reachability or error oracles, and can retrieve AWS metadata credentials where IMDSv1 accepts an unauthenticated GET, while IMDSv2 and GCP metadata prerequisites are not satisfied by the demonstrated request. This vulnerability is fixed in 0.31.0.

Record
CVE-2026-58196
CVSS 3.1
4.7 MediumCVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:L/I:N/A:N
Weakness
CWE-918
Updated
2026-11-02
EUVDEUVD-2026-78818 4.7

ToolHive is a utility designed to simplify the deployment and management of Model Context Protocol (MCP) servers. Prior to 0.31.0, remote.Handler.Authenticate in pkg/auth/remote/handler.go invokes discovery.DetectAuthenticationFromServer in pkg/auth/discovery/discovery.go, whose host-side HTTP clients trust remote-server-controlled authentication discovery destinations, follow redirects without host or scheme restrictions, and do not consistently block private addresses. A malicious or compromised remote MCP server can place a resource_metadata URL in WWW-Authenticate for ParseWWWAuthenticate to extract, after which FetchResourceMetadata, OIDC issuer discovery, and well-known discovery can issue GET requests to link-local, RFC1918, or other internal services outside the server's container. The user connects to a server that the user intends to use, but the server controls the internal destination; this path does not depend on the separate NAT64 guard omission because the affected clients did not invoke IsPrivateIP, and it differs from the DCR resolver path that already refused redirects. The demonstrated primitive reaches internal-only HTTP services and reachability or error oracles, and can retrieve AWS metadata credentials where IMDSv1 accepts an unauthenticated GET, while IMDSv2 and GCP metadata prerequisites are not satisfied by the demonstrated request. This vulnerability is fixed in 0.31.0.

Record
EUVD-2026-78818
CVSS 3.1
4.7 MediumCVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:L/I:N/A:N
Updated
2026-11-02
GitHub AdvisoriesGHSA-pr64-jmmf-jp54 2.9

ToolHive: SSRF in remote MCP server authentication discovery (host-side, bypasses container isolation)

Security Advisory: SSRF in remote MCP server authentication discovery

Severity: High. CWE: CWE-918. Affected: ToolHive through the latest release v0.29.3 and current main (HEAD b672d82f, 2026-06-12; re-verified 2026-06-14). FetchResourceMetadata and the discovery clients remain unguarded; no commits to pkg/auth/discovery or pkg/auth/remote address this. Originally identified at commit 05f11b53; all line references below are against HEAD b672d82f.

Summary

ToolHive's remote MCP server authentication discovery issues outbound HTTP requests to URLs the remote MCP server controls, with no private-IP or loopback guard and no restriction on redirects. ToolHive's core security model treats every MCP server as untrusted: the README states it "runs every MCP server in an isolated container" with "no local credentials," and it ships an egress proxy for network isolation. This discovery code runs host-side, in the ToolHive process, before and outside that per-server container sandbox. A malicious or compromised remote MCP server, added by a user through ToolHive's normal remote-server workflow, can therefore drive the ToolHive host itself to fetch arbitrary internal URLs, including cloud instance metadata, which bypasses the isolation ToolHive exists to provide. The user never selects a malicious target; they connect to a server they intend to use, and the attack is carried entirely in that server's discovery response.

ToolHive already establishes this boundary in code. ValidateRemoteURL (cmd/thv-operator/pkg/validation/url_validation.go:61) rejects internal IPs and known internal hostnames for the configured remote URL, and IsPrivateIP (pkg/networking/utilities.go:105) blocks RFC1918, link-local, 169.254.0.0/16, and loopback for outbound requests. The discovery clients below never call either guard, and the attacker-supplied resource_metadata URL and its redirect target are validated by neither. This is a deviation from the project's intended behavior, not a configuration choice by the operator.

The discovery clients are explicitly marked as trusted

The three outbound requests in this finding carry the maintainers' own gosec suppressions, each with a rationale comment asserting the URL is trusted:

  • discovery.go:165 -- resp, err := client.Do(req) // #nosec G704 -- targetURI is the MCP server endpoint URL from internal config
  • discovery.go:219 -- resp, err := client.Do(req) // #nosec G704 -- uri is built from the MCP server endpoint for auth discovery
  • discovery.go:931 -- resp, err := client.Do(req) // #nosec G704 -- URL is the OIDC well-known metadata endpoint

gosec's G704 is its SSRF taint-analysis rule: it flags exactly this pattern, an HTTP request whose URL flows from external input. The suppressions dismiss it on the premise that the URL comes "from internal config" or is "the MCP server endpoint." That premise is the bug. The targetURI is the remote MCP server endpoint, and the resource_metadata URL at line 931 comes straight from the server's WWW-Authenticate header (parsed by ParseWWWAuthenticate, discovery.go:293; resource_metadata extracted at discovery.go:315). Under ToolHive's own threat model the remote MCP server is untrusted, so "the MCP server endpoint" is attacker-controlled input, not internal config. A parallel cluster of //nolint:gosec // G706 (log-injection) suppressions on the same discovery responses (discovery.go:212, 221, 240, 268, 273, 280) shows the same data being treated as trusted throughout this code path. The maintainers saw the taint-analysis warning on these requests and waved it through on a trust assumption that contradicts the project's design.

Relationship to existing reports

This is distinct from issue #5135 (DCR resolver SSRF). #5135 is scoped to the DCR registration client, which already refuses redirects: client.CheckRedirect = errDCRRedirectRefused (pkg/auth/dcr/resolver.go:1281), with an in-code comment explaining that the…

Record
GHSA-pr64-jmmf-jp54
CVSS 4.0
2.9 LowCVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:N/VC:L/VI:N/VA:N/SC:N/SI:N/SA:N/E:P
CVSS 3.1
4.7 MediumCVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:L/I:N/A:N
Weakness
CWE-918
Updated
2026-11-02
NVD 4.7

ToolHive is a utility designed to simplify the deployment and management of Model Context Protocol (MCP) servers. Prior to 0.31.0, remote.Handler.Authenticate in pkg/auth/remote/handler.go invokes discovery.DetectAuthenticationFromServer in pkg/auth/discovery/discovery.go, whose host-side HTTP clients trust remote-server-controlled authentication discovery destinations, follow redirects without host or scheme restrictions, and do not consistently block private addresses. A malicious or compromised remote MCP server can place a resource_metadata URL in WWW-Authenticate for ParseWWWAuthenticate to extract, after which FetchResourceMetadata, OIDC issuer discovery, and well-known discovery can issue GET requests to link-local, RFC1918, or other internal services outside the server's container. The user connects to a server that the user intends to use, but the server controls the internal destination; this path does not depend on the separate NAT64 guard omission because the affected clients did not invoke IsPrivateIP, and it differs from the DCR resolver path that already refused redirects. The demonstrated primitive reaches internal-only HTTP services and reachability or error oracles, and can retrieve AWS metadata credentials where IMDSv1 accepts an unauthenticated GET, while IMDSv2 and GCP metadata prerequisites are not satisfied by the demonstrated request. This vulnerability is fixed in 0.31.0.

Record
CVE-2026-58196
CVSS 3.1
4.7 MediumCVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:L/I:N/A:N
Weakness
CWE-918
Updated
2026-11-02
OSV 4.7

ToolHive: SSRF in remote MCP server authentication discovery (host-side, bypasses container isolation)

ToolHive is a utility designed to simplify the deployment and management of Model Context Protocol (MCP) servers. Prior to 0.31.0, remote.Handler.Authenticate in pkg/auth/remote/handler.go invokes discovery.DetectAuthenticationFromServer in pkg/auth/discovery/discovery.go, whose host-side HTTP clients trust remote-server-controlled authentication discovery destinations, follow redirects without host or scheme restrictions, and do not consistently block private addresses. A malicious or compromised remote MCP server can place a resource_metadata URL in WWW-Authenticate for ParseWWWAuthenticate to extract, after which FetchResourceMetadata, OIDC issuer discovery, and well-known discovery can issue GET requests to link-local, RFC1918, or other internal services outside the server's container. The user connects to a server that the user intends to use, but the server controls the internal destination; this path does not depend on the separate NAT64 guard omission because the affected clients did not invoke IsPrivateIP, and it differs from the DCR resolver path that already refused redirects. The demonstrated primitive reaches internal-only HTTP services and reachability or error oracles, and can retrieve AWS metadata credentials where IMDSv1 accepts an unauthenticated GET, while IMDSv2 and GCP metadata prerequisites are not satisfied by the demonstrated request. This vulnerability is fixed in 0.31.0.

Record
CVE-2026-58196
CVSS 3.1
4.7 MediumCVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:L/I:N/A:N
Weakness
CWE-918
Updated
2026-11-02
OSVGHSA-pr64-jmmf-jp54 2.9

ToolHive: SSRF in remote MCP server authentication discovery (host-side, bypasses container isolation)

Security Advisory: SSRF in remote MCP server authentication discovery

Severity: High. CWE: CWE-918. Affected: ToolHive through the latest release v0.29.3 and current main (HEAD b672d82f, 2026-06-12; re-verified 2026-06-14). FetchResourceMetadata and the discovery clients remain unguarded; no commits to pkg/auth/discovery or pkg/auth/remote address this. Originally identified at commit 05f11b53; all line references below are against HEAD b672d82f.

Summary

ToolHive's remote MCP server authentication discovery issues outbound HTTP requests to URLs the remote MCP server controls, with no private-IP or loopback guard and no restriction on redirects. ToolHive's core security model treats every MCP server as untrusted: the README states it "runs every MCP server in an isolated container" with "no local credentials," and it ships an egress proxy for network isolation. This discovery code runs host-side, in the ToolHive process, before and outside that per-server container sandbox. A malicious or compromised remote MCP server, added by a user through ToolHive's normal remote-server workflow, can therefore drive the ToolHive host itself to fetch arbitrary internal URLs, including cloud instance metadata, which bypasses the isolation ToolHive exists to provide. The user never selects a malicious target; they connect to a server they intend to use, and the attack is carried entirely in that server's discovery response.

ToolHive already establishes this boundary in code. ValidateRemoteURL (cmd/thv-operator/pkg/validation/url_validation.go:61) rejects internal IPs and known internal hostnames for the configured remote URL, and IsPrivateIP (pkg/networking/utilities.go:105) blocks RFC1918, link-local, 169.254.0.0/16, and loopback for outbound requests. The discovery clients below never call either guard, and the attacker-supplied resource_metadata URL and its redirect target are validated by neither. This is a deviation from the project's intended behavior, not a configuration choice by the operator.

The discovery clients are explicitly marked as trusted

The three outbound requests in this finding carry the maintainers' own gosec suppressions, each with a rationale comment asserting the URL is trusted:

  • discovery.go:165 -- resp, err := client.Do(req) // #nosec G704 -- targetURI is the MCP server endpoint URL from internal config
  • discovery.go:219 -- resp, err := client.Do(req) // #nosec G704 -- uri is built from the MCP server endpoint for auth discovery
  • discovery.go:931 -- resp, err := client.Do(req) // #nosec G704 -- URL is the OIDC well-known metadata endpoint

gosec's G704 is its SSRF taint-analysis rule: it flags exactly this pattern, an HTTP request whose URL flows from external input. The suppressions dismiss it on the premise that the URL comes "from internal config" or is "the MCP server endpoint." That premise is the bug. The targetURI is the remote MCP server endpoint, and the resource_metadata URL at line 931 comes straight from the server's WWW-Authenticate header (parsed by ParseWWWAuthenticate, discovery.go:293; resource_metadata extracted at discovery.go:315). Under ToolHive's own threat model the remote MCP server is untrusted, so "the MCP server endpoint" is attacker-controlled input, not internal config. A parallel cluster of //nolint:gosec // G706 (log-injection) suppressions on the same discovery responses (discovery.go:212, 221, 240, 268, 273, 280) shows the same data being treated as trusted throughout this code path. The maintainers saw the taint-analysis warning on these requests and waved it through on a trust assumption that contradicts the project's design.

Relationship to existing reports

This is distinct from issue #5135 (DCR resolver SSRF). #5135 is scoped to the DCR registration client, which already refuses redirects: client.CheckRedirect = errDCRRedirectRefused (pkg/auth/dcr/resolver.go:1281), with an in-code comment explaining that the…

Record
GHSA-pr64-jmmf-jp54
CVSS 4.0
2.9 LowCVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:N/VC:L/VI:N/VA:N/SC:N/SI:N/SA:N/E:P
CVSS 3.1
4.7 MediumCVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:L/I:N/A:N
Weakness
CWE-918
Updated
2026-11-02
OSVGO-2026-6002 No score

ToolHive: SSRF in remote MCP server authentication discovery (host-side, bypasses container isolation) in github.com/stacklok/toolhive

ToolHive: SSRF in remote MCP server authentication discovery (host-side, bypasses container isolation) in github.com/stacklok/toolhive

Record
GO-2026-6002
Updated
2026-11-02
OSVopenSUSE-SU-2026:21483-1 No score

Security update for govulncheck-vulndb

This update for govulncheck-vulndb fixes the following issues:

  • Update to version 0.0.20260723T184607 2026-07-23T18:46:07Z (jsc#PED-11136).

Go CVE Numbering Authority IDs added or updated with aliases:

  • GO-2026-4730 CVE-2026-33022 GHSA-cv4x-93xx-wgfj
  • GO-2026-4761 CVE-2026-33211 GHSA-j5q5-j9gm-2w5c
  • GO-2026-5102 CVE-2026-45021 GHSA-3vcp-chfh-f6r2
  • GO-2026-5272 CVE-2026-40938 GHSA-94jr-7pqp-xhcq
  • GO-2026-5486 CVE-2026-40924 GHSA-m2cx-gpqf-qf74
  • GO-2026-5542 CVE-2026-40090 GHSA-pj97-4p9w-gx3q
  • GO-2026-5630 CVE-2026-25542 GHSA-rmx9-2pp3-xhcr
  • GO-2026-5643 CVE-2026-40923 GHSA-rx35-6rhx-7858
  • GO-2026-5675 CVE-2026-41492 GHSA-vvf7-6rmr-m29q
  • GO-2026-5711 CVE-2026-40161 GHSA-wjxp-xrpv-xpff
  • GO-2026-5746 CVE-2026-41567 GHSA-x86f-5xw2-fm2r
  • GO-2026-5749 CVE-2026-41328 GHSA-x92x-px7w-4gx4
  • GO-2026-5837 CVE-2026-44840 GHSA-q2m9-6jp9-c6mc
  • GO-2026-5929 CVE-2026-44342 GHSA-26v7-h57m-gh9m
  • GO-2026-5930 CVE-2026-33655 GHSA-6qcr-qxgr-m7fv
  • GO-2026-5933 GHSA-3g4q-2f67-2gvh
  • GO-2026-5934 GHSA-59qp-cfj3-rp64
  • GO-2026-5935 GHSA-7856-g3gv-9wq8
  • GO-2026-5936 CVE-2026-53552 GHSA-26rh-24rg-j3vv
  • GO-2026-5937 CVE-2026-53508 GHSA-2jcc-mxv7-p3f9
  • GO-2026-5938 CVE-2026-53553 GHSA-4g5x-hcwm-82jw
  • GO-2026-5939 CVE-2026-10517 GHSA-698x-9w2p-7vvp
  • GO-2026-5940 CVE-2026-53572 GHSA-6w3m-4hhp-775q
  • GO-2026-5941 CVE-2026-53487 GHSA-gvhc-wv3v-7pf8
  • GO-2026-5942 CVE-2026-46600
  • GO-2026-5944 CVE-2026-50197 GHSA-659f-rgp5-w4wf
  • GO-2026-5945 CVE-2026-6815 GHSA-rmxx-v9rj-vpvg
  • GO-2026-5946 CVE-2026-52831 GHSA-v5px-423j-pf7p
  • GO-2026-5947 CVE-2026-53649 GHSA-xqhv-chqm-fhcc
  • GO-2026-5948 CVE-2026-50554 GHSA-588f-fvcv-xhvf
  • GO-2026-5949 CVE-2026-50553 GHSA-rqrh-8wpv-x7hh
  • GO-2026-5950 CVE-2026-53602 GHSA-339v-266x-79xr
  • GO-2026-5951 CVE-2026-10219 GHSA-6jm8-4fhr-5w64
  • GO-2026-5952 CVE-2026-49834 GHSA-9vcr-p3rj-q5q6
  • GO-2026-5953 CVE-2026-9094 GHSA-c9w5-qp6m-m395
  • GO-2026-5956 CVE-2026-55252 GHSA-h5g6-xmh4-hc37
  • GO-2026-5957 CVE-2026-54158 GHSA-5xfx-xj4h-5p7r
  • GO-2026-5958 CVE-2026-54068 GHSA-gcm7-57gf-953c
  • GO-2026-5959 CVE-2026-54072 GHSA-h29v-hj44-q8cv
  • GO-2026-5961 CVE-2026-54069 GHSA-hvr9-72v2-fff3
  • GO-2026-5962 CVE-2026-54088 GHSA-m93h-4hw7-5qcm
  • GO-2026-5963 CVE-2026-54067 GHSA-mvjr-vv3c-w4qv
  • GO-2026-5964 CVE-2026-54066 GHSA-p4m3-mgmm-c664
  • GO-2026-5965 CVE-2026-54070 GHSA-w7cg-whh7-xp28
  • GO-2026-5966 CVE-2026-54089 GHSA-xqp3-jq6g-x3qm
  • GO-2026-5967 CVE-2026-50551 GHSA-56mp-4f3v-fgj2
  • GO-2026-5968 CVE-2026-54174 GHSA-fpg8-7664-jc5q
  • GO-2026-5969 GHSA-g936-7jqj-mwv8
  • GO-2026-5970 CVE-2026-56852
  • GO-2026-5973 CVE-2026-54250 GHSA-jxr7-mqhw-9p98
  • GO-2026-5974 CVE-2026-44300 GHSA-wmj8-9953-vff5
  • GO-2026-5975 CVE-2026-50018 GHSA-42j2-w334-qxw7
  • GO-2026-5976 CVE-2026-50141 GHSA-g7mm-9vx7-jm7h
  • GO-2026-5977 CVE-2026-50013 GHSA-qrh4-p6v4-mrfg
  • GO-2026-5978 CVE-2026-50125 GHSA-qw5r-ppcg-f8rj
  • GO-2026-5979 CVE-2026-50006 GHSA-xrcf-6jh3-ggvx
  • GO-2026-5980 CVE-2026-50158 GHSA-2c7f-fxww-6w6c
  • GO-2026-5981 GHSA-mqxv-9rm6-w8qc
  • GO-2026-5982 GHSA-pqg7-v6wh-3pfp
  • GO-2026-5984 CVE-2026-53604 GHSA-2p2f-px33-4vv5
  • GO-2026-5985 GHSA-7rx3-5wx3-5v76
  • GO-2026-5986 CVE-2026-61699 GHSA-cm26-5974-52h8
  • GO-2026-5987 CVE-2026-55513 GHSA-g4x6-jcvr-9m3g
  • GO-2026-5988 CVE-2026-54628 GHSA-hwrq-8wxh-q4xv
  • GO-2026-5989 CVE-2026-55512 GHSA-m3cx-mwpg-32jg
  • GO-2026-5990 CVE-2026-54629 GHSA-mf78-3rpf-r784
  • GO-2026-5991 CVE-2026-53603 GHSA-q4vm-pq3q-8wgq
  • GO-2026-5992 CVE-2026-61549 GHSA-qf34-295c-26v8
  • GO-2026-5996 CVE-2026-54450 GHSA-pph6-vfjv-vpjw
  • GO-2026-5997 CVE-2026-54452 GHSA-xgch-x3mx-cm3c
  • GO-2026-5998 CVE-2026-54495 GHSA-398h-7f66-3h4p
  • GO-2026-5999 CVE-2026-10814 GHSA-jh6h-v6mp-h22v
  • GO-2026-6002 CVE-2026-58196 GHSA-pr64-jmmf-jp54
  • GO-2026-6003 CVE-2026-53714 GHSA-22xc-xg2r-9j7v
  • GO-2026-6004 CVE-2026-52833 GHSA-3v79-m2cg-89ww
  • GO-2026-6005 CVE-2026-53715 GHSA-8fv2-88gg-hm7q
  • GO-2026-6006 CVE-2026-53716 GHSA-cxpq-8v7q-cg56 *…
Record
openSUSE-SU-2026:21483-1
Updated
2026-11-02

Affected products

Affected packages

Affected packages
PackageAffectedFixed
github.com/stacklok/toolhive
Go
 before 0.31.0
0.31.0
govulncheck-vulndb
openSUSE:Leap 16.0
From an advisory covering 1734 CVEs
 before 0.0.20260723T184607-160000.1.1
0.0.20260723T184607-160000.1.1

History

History
RecordedChangedSources
affected, facts
first indexedCVE List, EUVD, GitHub Advisories, NVD, OSV

Dates are when Every CVE recorded each change. Upstream dates may differ.

In short

CVE-2026-58196 is a medium vulnerability (CVSS 4.7) in Stacklok Toolhive, published July 15, 2026. No source lists public exploit code. EPSS puts the chance of exploitation in the next 30 days at 0.43%, the 37th percentile. Fixed in github.com/stacklok/toolhive 0.31.0.

References

Show 194 moreShow less
This page doesn't list everything

Left out here: 1,541 references and the rest of 3 source descriptions. The JSON record has them.

Pages All vulnerabilities Browse Docs Sources Guide Get API key
MoveEnter Open