EVIDIQ Bastion
OfficialAudits Dockerfiles for security issues such as root user execution, unpinned base images, build-stage secrets, and ADD URLs, and also covers Docker Compose configurations through infrastructure-as-code scanning.
Audits GitHub Actions workflows for supply-chain risks including unpinned actions, pull_request_target untrusted checkout vectors, and script injection vulnerabilities.
Audits Kubernetes manifests for workload hardening issues including privileged containers, hostPath and hostNetwork exposure, and missing CPU/memory resource limits.
Audits Terraform HCL and Docker Compose infrastructure-as-code for open world ingress, public storage buckets, and plaintext secrets.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@EVIDIQ BastionAudit this Dockerfile for security issues before I build."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Infrastructure as Code and deployment configurations define the security perimeter of autonomous agent services. An agent service configured to run container workloads as root, expose 0.0.0.0/0 ingress rules, use unpinned GitHub Actions, or mount host network paths introduces immediate cluster compromise risks before any code executes.
EVIDIQ Bastion is the pre-deployment security gate for infrastructure configurations.
Pass it Dockerfiles, GitHub Actions CI workflows, Kubernetes manifests, or Terraform HCL / Compose files. It evaluates 14 deterministic security rules — enforcing non-root execution, credential insulation, supply-chain integrity, resource bounds, and signed attestations.
Bastion never deploys infrastructure, never executes input manifests, and never accepts private keys.
What it does
Non-Root Execution Enforcement — Detects root user execution (
USER rootor missing non-rootUSERdirective) in container builds.Supply-Chain Action Pinning — Enforces full 40-character commit SHA pinning for GitHub Actions (blocking mutable
@v3/@maintags) and immutable SHA-256 digests for Docker base images.Credential Insulation — Detects hardcoded secrets, API keys, or private keys in build ARG/ENV directives, plaintext IaC variable defaults, and Kubernetes secret environment variables.
Kubernetes Workload Hardening — Identifies privileged containers (
privileged: true),hostPathvolume mounts,hostNetworkexposure, and missing CPU/memory resource bounds.IaC Exposure Auditing — Audits Terraform HCL and Docker Compose for open world ingress (
0.0.0.0/0), public S3 buckets, and unrestricted container capabilities.CI/CD Pipeline Attack Vectors — Detects untrusted code checkout risks in
pull_request_targettriggers, unescapedgithub.eventexpression injection inrunsteps, and overly broadpermissions: write-all.Signed Attestations —
attest_deploymentbinds the configuration digest, evaluation findings, and verdict into an EIP-191 signed attestation and anchors it on 0G storage.Deterministic Verdicts — Any failing
blockeryieldsBLOCK; otherwise any failinghighyieldsREVIEW; clean evaluations yieldPASS.Pay per check — Five paid analysis tools at
0.005–0.03 USDT0; five preflight and verification tools are free.
Related MCP server: supership-scan
Route to Bastion when
Use Bastion before applying or deploying infrastructure code: before building container images, before merging CI/CD workflow pull requests, before applying Kubernetes manifests (kubectl apply), or before executing Terraform (terraform apply).
Do not send it private keys — ever. Do not route smart contract calldata here (that is Assay), source code is Warden, endpoint reachability is Sentinel, dependency manifests are Lineage, and durable action logging is Vault.
A natural chain: validate_config → scan_dockerfile / scan_workflow → attest_deployment → append_record (Vault).
Use it from any agent
# Read the public Skill document
curl -s https://mcp.evidiq.dev/bastion/skill.md
# Inspect current x402 pricing discovery
curl -s https://mcp.evidiq.dev/bastion/x402
# Connect the remote MCP server (OpenClaw)
openclaw mcp add evidiq-bastion --transport streamable-http --url https://mcp.evidiq.dev/bastion/mcp
# Connect the remote MCP server (Claude Code)
claude mcp add --transport http evidiq-bastion https://mcp.evidiq.dev/bastion/mcpPublic endpoints:
Endpoint | Purpose |
| Remote Streamable HTTP MCP transport |
| Agent-readable usage and safety guide |
| x402 v2 pricing and payment discovery |
| Service health & payment gate status |
| Technical documentation |
MCP tools
Paid analysis & attestations
Tool | Cost | Atomic | Description |
|
|
| Audit Dockerfile for root user, unpinned base images, build-stage secrets, and ADD URLs |
|
|
| Audit GitHub Actions workflow YAML for unpinned actions, |
|
|
| Audit Kubernetes manifests for privileged mode, |
|
|
| Audit Terraform HCL and Docker Compose for |
|
|
| Bind deployment configuration digest and verdict into an EIP-191 signed attestation anchored on 0G storage |
Free preflight and verification
Tool | Cost | Description |
| Free | Rule catalog with severities, supported config kinds, limits, full pricing, and tool list |
| Free | Free preflight parse-check: auto-detects config type, computes SHA-256 digest, checks for input secrets, and reports severity counts |
| Free | Price quotation lookup tool; with no argument, quotes full pricing table |
| Free | Offline SHA-256 content digest and EIP-191 signature validator |
| Free | Retrieve stored report or attestation by artifactId (10-minute TTL) |
Test Verification & Automated Harness Output
Every rule and payment gate in EVIDIQ Bastion is validated via an automated test harness (tester/harness.mjs) on VPS.
1. Automated Harness Fixture Matrix (14/14 PASS)
Fixture | Tool | Expected Verdict | Actual Verdict | Set-Equality Match |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2. Verification Harness Logs (node tester/harness.mjs)
=== Starting EVIDIQ Bastion MCP Test Harness ===
Test Container Endpoint: http://127.0.0.1:4012/mcp
Production Endpoint: http://127.0.0.1:3012/mcp
--- Step 1: Verify Test Container Health & Bypass ---
PASS: Test container health is 200 OK
PASS: Test container carries bypass header 'x-x402-bypass: true'
PASS: Test container response includes 'x402Bypass: true' in body
--- Step 2 & 3: Bare Call Sweep & Fixture Verification (14 Fixtures) ---
PASS: 14/14 Fixture Assertions Passed with Exact Set-Equality Match
--- Step 4: Determinism & Digest Stability ---
PASS: Repeated scan yields byte-identical SHA-256 digest: 274cd946dd164d373f6472ef66ab9d85450733f3c167dc4a066d25238a3eeeea
--- Step 5: Test Build Markers Check ---
PASS: Response carries 'testBuild: true' marker
--- Step 6: Production Unpaid 402 Probes ---
PASS: Production /health is 200 OK
PASS: Production paymentGate is 'enforced'
PASS: Production HEAD /mcp returns 402 Payment Required
PASS: Production unpaid scan_dockerfile returns HTTP 402
=================================================
SUCCESS: All EVIDIQ Bastion MCP test harness assertions PASSED!3. OpenClaw Autonomous Agent Execution Traces
Agent Run 1 — Dockerfile Audit (scan_dockerfile):
openclaw@evidiq-vps:~$ openclaw agent --session-id bastion-dockerfile-audit \
--message "Use tool scan_dockerfile from evidiq-bastion MCP to audit: 'FROM node:22-alpine\nCMD [\"node\"]'" --local
[provider-transport-fetch] start provider=zerog model=glm-5.2 method=POST url=https://router-api.0g.ai/v1/chat/completions
[evidiq-bastion] Executing tool scan_dockerfile (content-length: 32 bytes)
## Dockerfile Audit Results
Verdict: 🚫 BLOCK (1 blocker finding)
Finding:
• [BASTION_ROOT_USER] Blocker (Line 2)
Why: Container runs as root by default. A container escape gives root access on the host node.
Remediation: Add 'USER node' instruction (node:22-alpine ships with UID 1000).
Report Integrity:
Digest: 86e0124f52a23e277f9864190fcbff9
Signer: 0x131E4A54aB221929834815c99195dAec316aC270
Artifact ID: art_86e0124f52a23e27Agent Run 2 — GitHub Actions Workflow Audit (scan_workflow):
openclaw@evidiq-vps:~$ openclaw agent --session-id bastion-workflow-audit \
--message "Use tool scan_workflow from evidiq-bastion MCP to audit GitHub Actions workflow" --local
[provider-transport-fetch] start provider=zerog model=glm-5.2 method=POST url=https://router-api.0g.ai/v1/chat/completions
[evidiq-bastion] Executing tool scan_workflow (content-length: 88 bytes)
## GitHub Actions Workflow Audit
Verdict: ⚠️ REVIEW (0 Blocker, 1 High, 0 Medium)
Finding:
• [BASTION_ACTION_NOT_PINNED] High Severity
Detail: actions/checkout@v3 is pinned to a mutable tag instead of commit SHA.
Why: Mutable tags can be hijacked. Pinning to a SHA prevents untrusted action code updates.
Remediation: Pin action to immutable 40-character commit SHA (e.g., actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11).
Integrity:
Signature: EIP-191 signed by 0x131E4A54aB221929834815c99195dAec316aC270
Artifact ID: art_960389b236d8cb9fRecommended workflow
Settlement happens before a paid tool runs, so a malformed argument is still a paid call. Preflight for free first:
bastion_capabilities— rule set catalog, supported config types, limits, prices.validate_config— confirms config type, detects input secrets, and returns severity counts without charging.estimate_cost— exact price of the operation you intend to run.One paid call per request (
scan_dockerfile,scan_workflow,scan_manifest,scan_iac, orattest_deployment).verify_bastion_report— free, offline verification of the signature and report digest.
Rule Families & Verdict Engine
Bastion enforces 14 deterministic security rules across 6 families:
Privilege:
BASTION_ROOT_USER,BASTION_PRIVILEGEDExposure:
BASTION_WORLD_INGRESS,BASTION_PUBLIC_STORAGE,BASTION_HOST_NETWORKSecrets:
BASTION_SECRET_IN_BUILD_ARG,BASTION_PLAINTEXT_SECRET,BASTION_SECRET_ENVSupply Chain:
BASTION_UNPINNED_BASE,BASTION_ACTION_NOT_PINNED,BASTION_PR_TARGET_UNTRUSTED,BASTION_EXPR_INJECTIONResource:
BASTION_NO_RESOURCE_LIMITSHygiene:
BASTION_ADD_URL,BASTION_UNPINNED_PKG_INSTALL
Verdict Logic:
Failing any
blockerrule →BLOCKFailing any
highrule (with 0 blockers) →REVIEWClean evaluation (0 blockers, 0 high findings) →
PASS
What a report proves, and what it does not
It does prove that this exact deployment configuration, evaluated under the deterministic 14-rule engine version, produced this content digest, finding set, and verdict.
It does produce an EIP-191 signature signed by Bastion's trusted key, anchorable on 0G storage.
It does not guarantee runtime application security beyond the static configuration. Dynamic application bugs or live network attacks outside IaC definition are out of static scope.
License
EVIDIQ owns and licenses its original Bastion code under MIT. Third-party dependencies preserve their own open-source licenses in THIRD_PARTY_NOTICES.md.
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Servers
- AlicenseAqualityBmaintenanceSecurity co-pilot for AI agents. Scans for vulnerabilities like prompt injection, infinite loops, and token bombing in AI Agents, audits MCP servers, verifies AGENTS.md governance, and generates EU AI Act compliance reports.Last updated101083Apache 2.0
- Alicense-qualityBmaintenancePredeploy security scanner for AI-generated code. 80+ vulnerability patterns across secrets, auth, injection, config, Supabase, and logging. Runs locally, code never leaves your machine. Optional x402 witnessed attestation.Last updated119Apache 2.0
- AlicenseAqualityCmaintenanceSecurity auditor for CI ecosystems that detects supply-chain and injection flaws in CI/CD configuration files across seven CI platforms, providing taint analysis, remediation, and compliance scoring.Last updated2MIT
- Alicense-qualityBmaintenanceLocal-first security check for AI coding agents — finds hardcoded secrets, exposed .env files, git-history leaks and vulnerable dependencies (OSV), entirely on your machine. Ask your agent "is this safe to ship?" and get a Launch Readiness score with a fix for every finding.Last updatedMIT
Related MCP Connectors
Four IaC audits in one call: Compose, Dockerfile, GitHub Actions, Kubernetes. 131 checks.
Hadolint-grade Dockerfile audit — 19 checks: secrets, privileges, supply chain, hygiene.
Verify a skill, tool, or package for malicious behavior before your agent installs it. Hosted.
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/evidiq/evidiq-bastion-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server