{"id":"CVE-2026-46519","aliases":["EUVD-2026-36286","GHSA-cr22-wjx7-2w6m","openSUSE-SU-2026:10914-1","openSUSE-SU-2026:21152-1"],"status":"published","title":"mcp-server-kubernetes Affected By Tool Access Control Bypass: Presentation-Layer Filtering Without Execution-Layer Enforcement","description":"mcp-server-kubernetes is a Model Context Protocol server for Kubernetes cluster management. Prior to version 3.6.0, mcp-server-kubernetes exposes three environment variables (ALLOW_ONLY_READONLY_TOOLS, ALLOW_ONLY_NON_DESTRUCTIVE_TOOLS, ALLOWED_TOOLS) documented as access controls for restricting which Kubernetes operations are available. These controls are enforced at the tool discovery layer (tools/list) but not at the execution layer (tools/call). Any client that knows a tool name can invoke it directly regardless of the configured restriction mode. The access control was effectively cosmetic. This issue has been patched in version 3.6.0.","severity":{"version":"3.1","score":8.8,"vector":"CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H","level":"high","assessor":"cna","from":"cve"},"epss":{"day":"2026-11-11","score":0.00513,"percentile":0.4421,"model":"v2026.06.15"},"facts":{"in_kev":false,"known_exploited":false,"public_exploit":false,"fix_available":true},"cwes":["CWE-863"],"published":"2026-05-21T20:33:46Z","modified":"2026-10-28T02:31:05Z","fixes":[{"ecosystem":"npm","package":"mcp-server-kubernetes","fixed_in":["3.6.0"],"sources":["ghsa","osv"]},{"ecosystem":"openSUSE:Tumbleweed","package":"atril","fixed_in":["1.28.4-1.1"],"sources":["osv"]},{"ecosystem":"openSUSE:Leap 16.0","package":"atril","fixed_in":["1.28.4-bp160.1.1"],"sources":["osv"]}],"affected":[{"source":"ghsa","ecosystem":"npm","package":"mcp-server-kubernetes","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"3.6.0"}]}]},{"source":"osv","ecosystem":"npm","package":"mcp-server-kubernetes","ranges":[{"type":"SEMVER","events":[{"introduced":"0"},{"fixed":"3.6.0"}]}]},{"source":"osv","ecosystem":"openSUSE:Tumbleweed","package":"atril","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"1.28.4-1.1"}]}]},{"source":"osv","ecosystem":"openSUSE:Leap 16.0","package":"atril","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"1.28.4-bp160.1.1"}]}]}],"products":[{"vendor":"flux159","product":"mcp-server-kubernetes"}],"references":[{"url":"https://github.com/Flux159/mcp-server-kubernetes/security/advisories/GHSA-cr22-wjx7-2w6m","tags":["x_refsource_CONFIRM","exploit","advisory","web"]},{"url":"https://github.com/Flux159/mcp-server-kubernetes/releases/tag/v3.6.0","tags":["x_refsource_MISC","web"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-46519","tags":["advisory"]},{"url":"https://github.com/advisories/GHSA-cr22-wjx7-2w6m"},{"url":"https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/46xxx/CVE-2026-46519.json","tags":["advisory"]},{"url":"https://github.com/Flux159/mcp-server-kubernetes","tags":["package"]},{"url":"","tags":["advisory"]},{"url":"https://www.suse.com/security/cve/CVE-2026-46519","tags":["web"]},{"url":"https://bugzilla.suse.com/1265880","tags":["report"]}],"sources":{"cve":[{"id":"CVE-2026-46519","title":"mcp-server-kubernetes Affected By Tool Access Control Bypass: Presentation-Layer Filtering Without Execution-Layer Enforcement","description":"mcp-server-kubernetes is a Model Context Protocol server for Kubernetes cluster management. Prior to version 3.6.0, mcp-server-kubernetes exposes three environment variables (ALLOW_ONLY_READONLY_TOOLS, ALLOW_ONLY_NON_DESTRUCTIVE_TOOLS, ALLOWED_TOOLS) documented as access controls for restricting which Kubernetes operations are available. These controls are enforced at the tool discovery layer (tools/list) but not at the execution layer (tools/call). Any client that knows a tool name can invoke it directly regardless of the configured restriction mode. The access control was effectively cosmetic. This issue has been patched in version 3.6.0.","status":"published","published":"2026-06-11T18:34:15.281Z","modified":"2026-06-11T19:40:22.771Z","cvss":[{"version":"3.1","score":8.8,"vector":"CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H","level":"high","assessor":"cna"}],"cwes":["CWE-863"],"updated_at":"2026-11-02T13:46:55.086623Z"}],"euvd":[{"id":"EUVD-2026-36286","description":"mcp-server-kubernetes is a Model Context Protocol server for Kubernetes cluster management. Prior to version 3.6.0, mcp-server-kubernetes exposes three environment variables (ALLOW_ONLY_READONLY_TOOLS, ALLOW_ONLY_NON_DESTRUCTIVE_TOOLS, ALLOWED_TOOLS) documented as access controls for restricting which Kubernetes operations are available. These controls are enforced at the tool discovery layer (tools/list) but not at the execution layer (tools/call). Any client that knows a tool name can invoke it directly regardless of the configured restriction mode. The access control was effectively cosmetic. This issue has been patched in version 3.6.0.","status":"published","published":"2026-06-11T18:34:15Z","modified":"2026-06-11T19:40:22Z","cvss":[{"version":"3.1","score":8.8,"vector":"CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H","level":"high"}],"updated_at":"2026-11-02T18:20:44.061211Z"}],"ghsa":[{"id":"GHSA-cr22-wjx7-2w6m","title":"MCP Server Kubernetes: Tool Access Control Bypass via Presentation-Layer Filtering Without Execution-Layer Enforcement","description":"## Summary\n\n`mcp-server-kubernetes` exposes three environment variables (`ALLOW_ONLY_READONLY_TOOLS`, `ALLOW_ONLY_NON_DESTRUCTIVE_TOOLS`, `ALLOWED_TOOLS`) documented as access controls for restricting which Kubernetes operations are available. These controls are enforced at the tool discovery layer (`tools/list`) but not at the execution layer (`tools/call`). Any client that knows a tool name can invoke it directly regardless of the configured restriction mode. The access control was effectively cosmetic.\n\nFixed in v3.6.0.\n\n## Impact\n\nAn attacker or misconfigured AI agent with network access to the MCP server's HTTP endpoint could invoke any Kubernetes tool regardless of the restriction mode configured by the operator -- including `kubectl_delete`, `exec_in_pod`, `kubectl_generic`, and `node_management`.\n\nThe project explicitly supports and documents multi-client HTTP deployment scenarios (Streamable HTTP and SSE transports, in-cluster deployments, Codex CLI and Gemini CLI integrations). In these deployments, operators relied on the tool restriction env vars to enforce least-privilege access across users or roles. The bypass invalidated that model entirely.\n\nSeverity scales with the Kubernetes service account's permissions. In environments where the MCP server runs with `cluster-admin` (common in dev/staging), this is equivalent to full cluster compromise for any client that can reach the endpoint.\n\nThe `MCP_AUTH_TOKEN` / `X-MCP-AUTH` mechanism controls who can reach the endpoint but provides no per-tool authorization. An authenticated client restricted to `ALLOWED_TOOLS=kubectl_get` could still invoke `kubectl_delete` after authentication.\n\n## Root Cause\n\nIn `src/index.ts`, the `ListToolsRequestSchema` handler applied the configured filtering logic before returning available tools. The `CallToolRequestSchema` handler dispatched directly by tool name with no equivalent check -- every tool was reachable unconditionally.\n\n## Proof of Concept\n\nTested across all three restriction modes against a live kind cluster. In each case, `kubectl_delete` was absent from `tools/list` but executed successfully via a direct `tools/call` request:\n\n```shell\ncurl -s http://\u003cHOST\u003e:3003/mcp \\\n  -H 'Content-Type: application/json' \\\n  -H 'Accept: application/json, text/event-stream' \\\n  -d '{\"jsonrpc\":\"2.0\",\"id\":2,\"method\":\"tools/call\",\"params\":{\"name\":\"kubectl_delete\",\"arguments\":{\"resourceType\":\"pod\",\"name\":\"test-pod\",\"namespace\":\"default\"}}}'\n```\n\nResult: `{\"result\":{\"content\":[{\"type\":\"text\",\"text\":\"pod \\\"test-pod\\\" deleted\\n\"}]}}`\n\nConfirmed across `ALLOW_ONLY_READONLY_TOOLS=true`, `ALLOW_ONLY_NON_DESTRUCTIVE_TOOLS=true`, and `ALLOWED_TOOLS=kubectl_get`.\n\n## Remediation\n\nThe fix applies the same filtering logic from `ListToolsRequestSchema` at the start of the `CallToolRequestSchema` handler, returning an error for any tool call outside the active allowed set. Fixed in v3.6.0.\n\n## Credit\n\nDiscovered by [Francisco Rosales](https://www.linkedin.com/in/francisco-rosales-celis/) of [Manifold Security](https://manifold.security), coordinated by [Ax Sharma](https://www.linkedin.com/in/axsharma/), Head of Research at Manifold Security.","status":"published","published":"2026-05-21T20:33:46Z","modified":"2026-06-12T19:26:32Z","cvss":[{"version":"3.1","score":8.8,"vector":"CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H","level":"high"}],"cwes":["CWE-863"],"updated_at":"2026-11-02T14:40:27.329706Z"}],"nvd":[{"id":"CVE-2026-46519","description":"mcp-server-kubernetes is a Model Context Protocol server for Kubernetes cluster management. Prior to version 3.6.0, mcp-server-kubernetes exposes three environment variables (ALLOW_ONLY_READONLY_TOOLS, ALLOW_ONLY_NON_DESTRUCTIVE_TOOLS, ALLOWED_TOOLS) documented as access controls for restricting which Kubernetes operations are available. These controls are enforced at the tool discovery layer (tools/list) but not at the execution layer (tools/call). Any client that knows a tool name can invoke it directly regardless of the configured restriction mode. The access control was effectively cosmetic. This issue has been patched in version 3.6.0.","status":"published","published":"2026-06-11T19:16:42.213Z","modified":"2026-10-26T11:10:00.157Z","cvss":[{"version":"3.1","score":8.8,"vector":"CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H","level":"high","assessor":"secondary"}],"cwes":["CWE-863"],"updated_at":"2026-11-02T14:12:23.485267Z"}],"osv":[{"id":"CVE-2026-46519","title":"mcp-server-kubernetes Affected By Tool Access Control Bypass: Presentation-Layer Filtering Without Execution-Layer Enforcement","description":"mcp-server-kubernetes is a Model Context Protocol server for Kubernetes cluster management. Prior to version 3.6.0, mcp-server-kubernetes exposes three environment variables (ALLOW_ONLY_READONLY_TOOLS, ALLOW_ONLY_NON_DESTRUCTIVE_TOOLS, ALLOWED_TOOLS) documented as access controls for restricting which Kubernetes operations are available. These controls are enforced at the tool discovery layer (tools/list) but not at the execution layer (tools/call). Any client that knows a tool name can invoke it directly regardless of the configured restriction mode. The access control was effectively cosmetic. This issue has been patched in version 3.6.0.","status":"published","published":"2026-06-11T18:34:15.281Z","modified":"2026-10-28T02:31:05.155248252Z","cvss":[{"version":"3.1","score":8.8,"vector":"CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H","level":"high"}],"cwes":["CWE-863"],"updated_at":"2026-11-02T15:45:05.870292Z"},{"id":"GHSA-cr22-wjx7-2w6m","title":"MCP Server Kubernetes: Tool Access Control Bypass via Presentation-Layer Filtering Without Execution-Layer Enforcement","description":"## Summary\n\n`mcp-server-kubernetes` exposes three environment variables (`ALLOW_ONLY_READONLY_TOOLS`, `ALLOW_ONLY_NON_DESTRUCTIVE_TOOLS`, `ALLOWED_TOOLS`) documented as access controls for restricting which Kubernetes operations are available. These controls are enforced at the tool discovery layer (`tools/list`) but not at the execution layer (`tools/call`). Any client that knows a tool name can invoke it directly regardless of the configured restriction mode. The access control was effectively cosmetic.\n\nFixed in v3.6.0.\n\n## Impact\n\nAn attacker or misconfigured AI agent with network access to the MCP server's HTTP endpoint could invoke any Kubernetes tool regardless of the restriction mode configured by the operator -- including `kubectl_delete`, `exec_in_pod`, `kubectl_generic`, and `node_management`.\n\nThe project explicitly supports and documents multi-client HTTP deployment scenarios (Streamable HTTP and SSE transports, in-cluster deployments, Codex CLI and Gemini CLI integrations). In these deployments, operators relied on the tool restriction env vars to enforce least-privilege access across users or roles. The bypass invalidated that model entirely.\n\nSeverity scales with the Kubernetes service account's permissions. In environments where the MCP server runs with `cluster-admin` (common in dev/staging), this is equivalent to full cluster compromise for any client that can reach the endpoint.\n\nThe `MCP_AUTH_TOKEN` / `X-MCP-AUTH` mechanism controls who can reach the endpoint but provides no per-tool authorization. An authenticated client restricted to `ALLOWED_TOOLS=kubectl_get` could still invoke `kubectl_delete` after authentication.\n\n## Root Cause\n\nIn `src/index.ts`, the `ListToolsRequestSchema` handler applied the configured filtering logic before returning available tools. The `CallToolRequestSchema` handler dispatched directly by tool name with no equivalent check -- every tool was reachable unconditionally.\n\n## Proof of Concept\n\nTested across all three restriction modes against a live kind cluster. In each case, `kubectl_delete` was absent from `tools/list` but executed successfully via a direct `tools/call` request:\n\n```shell\ncurl -s http://\u003cHOST\u003e:3003/mcp \\\n  -H 'Content-Type: application/json' \\\n  -H 'Accept: application/json, text/event-stream' \\\n  -d '{\"jsonrpc\":\"2.0\",\"id\":2,\"method\":\"tools/call\",\"params\":{\"name\":\"kubectl_delete\",\"arguments\":{\"resourceType\":\"pod\",\"name\":\"test-pod\",\"namespace\":\"default\"}}}'\n```\n\nResult: `{\"result\":{\"content\":[{\"type\":\"text\",\"text\":\"pod \\\"test-pod\\\" deleted\\n\"}]}}`\n\nConfirmed across `ALLOW_ONLY_READONLY_TOOLS=true`, `ALLOW_ONLY_NON_DESTRUCTIVE_TOOLS=true`, and `ALLOWED_TOOLS=kubectl_get`.\n\n## Remediation\n\nThe fix applies the same filtering logic from `ListToolsRequestSchema` at the start of the `CallToolRequestSchema` handler, returning an error for any tool call outside the active allowed set. Fixed in v3.6.0.\n\n## Credit\n\nDiscovered by [Francisco Rosales](https://www.linkedin.com/in/francisco-rosales-celis/) of [Manifold Security](https://manifold.security), coordinated by [Ax Sharma](https://www.linkedin.com/in/axsharma/), Head of Research at Manifold Security.","status":"published","published":"2026-05-21T20:33:46Z","modified":"2026-06-12T19:30:11.497746088Z","cvss":[{"version":"3.1","score":8.8,"vector":"CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H","level":"high"}],"cwes":["CWE-863"],"updated_at":"2026-11-02T15:46:45.28161Z"},{"id":"openSUSE-SU-2026:10914-1","title":"atril-1.28.4-1.1 on GA media","description":"These are all security issues fixed in the atril-1.28.4-1.1 package on the GA media of openSUSE Tumbleweed.","status":"published","published":"2026-05-31T00:00:00Z","modified":"2026-06-01T07:45:13.400782125Z","updated_at":"2026-11-02T16:58:38.95195Z"},{"id":"openSUSE-SU-2026:21152-1","title":"Security update for atril","description":"This update for atril fixes the following issues:\n\nChanges in atril:\n\n- Update to version 1.28.4 (bsc#1265880 CVE-2026-46519):\n  * Build fixes\n  * Fix tests imported from XReader\n  * Fix tests with AT-SPI2 \u003e= 2.53\n  * Improve search system\n  * pdf: Always use poppler_document_save to avoid data loss\n  * Use properties for can-zoom-in and -out\n  * libview: Allow zooming to the limits of the scale\n  * shell: Fix Max zoom in UI\n\n- Update to version 1.28.2:\n  * epub: Disable thumbnailing sidebar\n  * Fix .cbr mimetype\n  * Wayland: stop segfaults on some systems\n  * replace deprecated gtk_menu_tool_button_new_from_stock\n  * libview/ev-document-model.c remove one more deprecation warning\n  * replace ev_document_model_get_dual_page with\n\n- Update to version 1.28.1:\n  * Update CBR library to libarchive in README.md\n  * ci: fix travis build failures caused by recent travis changes\n  * Cleanup icons Makefile\n  * icons: Include higher resolution icons\n  * Updated translations.\n","status":"published","published":"2026-06-23T13:25:31Z","modified":"2026-06-30T18:24:45.662416259Z","updated_at":"2026-11-02T16:58:41.885092Z"}]}}