# rad-security/mcp-server [Health: Active]

**Category:** 🔒 Security  
**Repository:** https://github.com/rad-security/mcp-server  
**GitHub Stars:** 6  
**npm Downloads (last month):** 822  
**Views:** 3  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/rad-security-mcp-server

## Description
MCP server for RAD Security, providing AI-powered security insights for Kubernetes and cloud environments. This server provides tools for querying the Rad Security API and retrieving security findings, reports, runtime data and many more.

## Tools
Capabilities this server exposes over MCP:

- **list_containers** — List containers secured by RAD Security with optional filtering by image name, image digest, namespace, cluster_id, or free text search
- **get_container_details** — Get detailed information about a container secured by RAD Security
- **list_clusters** — List Kubernetes clusters managed by RAD Security
- **get_cluster_details** — Get detailed information about a specific Kubernetes cluster managed by RAD Security
- **who_shelled_into_pod** — Get k8s audit logs with information about users who shelled into a pod
- **list_images** — List container images with optional filtering by page, page size, sort, and search query
- **list_image_vulnerabilities** — List vulnerabilities in a container image with optional filtering by severity
- **get_top_vulnerable_images** — Get the most vulnerable images from your account
- **get_image_sbom** — Get the SBOM of a container image
- **ignore_cve** — Ignore a CVE for this account so it no longer appears in vulnerability reporting. Use for confirmed false positives, accepted risks, or won't-fix decisions. Do NOT use for remediated CVEs — those drop off automatically on the next scan.
- **unignore_cve** — Remove an account-wide CVE disposition, restoring the CVE to vulnerability reporting.
- **list_cve_dispositions** — List active CVE dispositions (ignored / false positive) for this account, with reason and author.
- **get_k8s_resource_details** — Get the latest manifest of a Kubernetes resource
- **list_k8s_resources** — List Kubernetes resources with optional filtering by namespace, resource types, and cluster
- **get_containers_process_trees** — Get process trees for multiple containers
- **get_containers_baselines** — Get runtime baselines for multiple containers
- **get_container_llm_analysis** — Get LLM analysis of a container's process tree
- **list_security_findings** — List security findings with optional filtering by types, severities, sources, and status
- **update_security_finding_status** — Update the status of a security finding
- **mark_inbox_item_as_false_positive** — Mark an inbox item as a false positive with a reason
- **list_inbox_items** — List inbox items with optional filtering by any field. Multiple filters can be combined eg. 'search:cve-2024-12345 and severity:high'
- **get_inbox_item_details** — Get detailed information about a specific inbox item
- **list_workflows** — List all workflows
- **get_workflow** — Get detailed information about a specific workflow by ID. It contains the workflow definition, default arguments, and schema how to run the workflow
- **list_workflow_runs** — List workflow runs with optional filtering by workflow ID
- **get_workflow_run** — Get detailed information about a specific workflow run
- **run_workflow** — Run a workflow with optional argument overrides
- **list_workflow_schedules** — List workflow schedules with optional filtering by workflow ID
- **create_custom_workflow** — Create a new automation (a Windmill workflow) from a YAML definition. Pass the YAML document itself as a string, not a file path. It is validated server-side before deployment; on failure nothing is deployed and the errors are returned. Returns the new automation WITHOUT echoing the definition back — use `id` from the result when referring to it, and `get_workflow` if you need to read the definition.
- **update_custom_workflow** — Update an existing automation with new YAML. Only automations created via create_custom_workflow can be updated. Returns the updated automation without echoing the definition back.
- **add_workflow_schedule** — Add a cron-based schedule to an automation so it runs automatically at the specified times. This ADDS a schedule; to change when an automation runs, use update_workflow_schedule instead, or the automation will fire at both the old time and the new one.
- **update_workflow_schedule** — Change an existing schedule's cron expression or timezone. Use this to move when an automation runs. Get the schedule_id from list_workflow_schedules.
- **delete_workflow_schedule** — Remove a schedule from an automation so it stops running on that cadence. The automation itself is not deleted. Get the schedule_id from list_workflow_schedules.
- **search_knowledge_base** — Search your organization's knowledge base to find relevant uploaded documents, procedures, reports, and other content using natural language queries
- **list_knowledge_base_collections** — List all collections in your organization's knowledge base. Collections are used to organize and categorize documents
- **list_knowledge_base_documents** — List documents in your organization's knowledge base with optional filtering by collections, file type, or status
- **query_knowledge_base_document** — Query a CSV document from the knowledge base using natural language. IMPORTANT: This tool ONLY works with CSV documents. Use list_knowledge_base_documents with filters='file_type:csv' to find CSV document IDs (search_knowledge_base results also contain document IDs). Results are returned as a markdown table
- **get_knowledge_base_document_content** — Get the FULL text content of a knowledge base document (extracted text for PDF/DOCX, the raw file for markdown/plaintext/CSV). Use this to read or analyze a whole document rather than the excerpts search_knowledge_base returns. Find document IDs via list_knowledge_base_documents or search_knowledge_base results
- **get_knowledge_base_document_download_url** — Get a time-limited download URL for the ORIGINAL document file (any format, including PDF/DOCX binaries). Use when you need the original file itself — e.g. to fetch it into a sandbox for structural parsing (tables, layout), or when get_knowledge_base_document_content reports no text available. For reading text, prefer get_knowledge_base_document_content
- **radql_list_data_types** — List all available RadQL data types (discovery). ALWAYS call this FIRST before using other RadQL tools to discover what data is available to query. Returns data types like 'containers', 'kubernetes_resources', 'inbox_items', 'cloud_resources', 'cloud_benchmarks', 'cloud_benchmark_summaries', etc. with descriptions.
- **radql_get_type_metadata** — Get schema/metadata for a specific RadQL data type. Shows available fields, data types, which fields can be filtered/searched, and provides query examples. Call this AFTER radql_list_data_types to understand how to query a specific data type.
- **radql_list_filter_values** — List possible values for a filter field (e.g., namespace list, cluster list, severity values). Useful for building dynamic filters when you need to know available enum-like values. Call this when constructing filters that need specific values.
- **radql_query** — Execute RadQL queries for security investigations. Supports: list (filter/search), get_by_id (single item), stats (aggregations).

WORKFLOW: radql_list_data_types -> radql_get_type_metadata -> radql_query

COMMON FIELDS BY DATA TYPE:
containers: name, image_name, image_repo, owner_kind, cluster_id, created_at
  Example: image_name:*nginx* AND owner_kind:Pod

finding_groups: type, source_kind, source_name, rule_title, severity, event_timestamp
  Types: k8s_misconfiguration, k8s_audit_logs_anomaly
  Example: type:k8s_misconfiguration AND severity:critical

inbox_items: severity (High|Medium|Low), type, title, archived, false_positive, created_at
  Example: severity:High AND archived:false

kubernetes_resources: kind, name, namespace, cluster_id, owner_kind, created_at
  Example: kind:Deployment AND namespace:production

CLOUD RESOURCES & COMPLIANCE (use these RadQL data types instead of dedicated cloud tools):
cloud_resources: cloud_provider, cloud_account_id, resource_type, resource_name, resource_id, resource_json, last_seen_at
  Example: cloud_provider:aws AND resource_type:aws_iam_policy

cloud_benchmark_summaries: cloud_provider, cloud_account_id, benchmark_id, title, description, fail_count, pass_count, total_count, last_seen_at
  Example: cloud_provider:aws AND fail_count>0

cloud_benchmarks: cloud_provider, cloud_account_id, benchmark_id, control_id, control_title, severity, status, reason, resource_id, last_seen_at
  Example: status:fail AND benchmark_id:*cis*

CRITICAL QUOTING RULES:
MUST quote when value contains:
  - Dates/timestamps: created_at>"2024-01-01" (NOT created_at>2024-01-01)
  - Hyphens: cluster_id:"abc-123-def", name:"kube-system"
  - UUIDs: id:"550e8400-e29b-41d4-a716-446655440000"
  - Spaces: title:"my alert"
  - Special chars: :, =, <, >, !, (, )
  - Wildcards with hyphens: name:"kube-*"

OK to leave unquoted:
  - Simple strings: status:active, kind:Pod
  - Numbers: count:123
  - Booleans: archived:true
  - Simple wildcards: name:nginx*

For complete schema: call radql_get_type_metadata with target data_type
- **radql_query_builder** — Helper tool to build RadQL queries programmatically from structured conditions. Useful when you need to construct complex filter or stats queries from structured inputs.
- **radql_batch_query** — Execute multiple RadQL queries in parallel for efficiency. Useful for fetching related data from different data types simultaneously (e.g., container details + vulnerabilities + network connections).
- **list_widget_templates** — List widget templates with optional filtering by visualization type and category
- **get_widget_template** — Get detailed information about a specific widget template
- **list_dashboard_templates** — Browse the catalogue of ready-made dashboards, each with working queries for a given area (containers, images, runtime, compliance, findings). Returns id, title, description and category WITHOUT the rows — call get_dashboard_template with the id to read a definition and copy its widgets and queries.
- **get_dashboard_template** — Get detailed information about a specific dashboard template
- **list_dashboards** — List dashboards for the account
- **get_dashboard** — Get detailed information about a specific dashboard
- **create_dashboard** — Create a dashboard for the account. Build `rows` from the widget templates (list_widget_templates / get_widget_template) so the visualization and query shapes are valid.
- **update_dashboard** — Update an existing dashboard. Omitted fields are left unchanged, so a small edit (a title, one row) does not require resending the whole dashboard.
- **list_external_integrations** — List external integrations configured for the tenant (e.g., Slack, AWS CloudTrail, Okta). Returns integration details including capabilities, configuration, mcp support and sync status.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "mcp-server": {
    "command": "npx",
    "args": ["-y","@rad-security/mcp-server"]
  }
}
```

## Documentation

## What rad-security/mcp-server MCP server does

The rad-security/mcp-server MCP server exposes RAD Security account data and actions to MCP-compatible AI clients. Its tools cover Kubernetes clusters and resources, secured containers, container images, vulnerabilities, SBOMs, runtime process data, security findings, inbox items, pod shell access, and workflow automations.

Inventory tools list clusters, containers, images, and Kubernetes resources, with filters available for several resource types. Investigation tools retrieve container and cluster details, image vulnerabilities, SBOMs, process trees, runtime baselines, LLM analysis of container processes, audit records, findings, and inbox item details.

The server also includes controlled write operations. An agent can change a finding status, mark an inbox item as a false positive, ignore or restore a CVE disposition, run a workflow, or create and update custom Windmill workflows from YAML. Ignoring a CVE is intended for false positives, accepted risks, or won't-fix decisions; remediated CVEs are expected to disappear after a later scan instead.

## How it works

The hosted service is available at `https://api.rad.security/mcp/` over Streamable HTTP. The trailing slash is required. Each request is authenticated against the RAD Security API with an `Authorization` bearer value containing an access key ID, secret key, and account ID.

By default, a connection receives every enabled toolkit. Request headers can reduce the exposed surface: `X-Rad-Toolkits` selects toolkits, `X-Rad-Exclude-Toolkits` removes selected toolkits, and `X-Rad-Readonly: true` limits the connection to read-only tools. This allows a client configuration to separate investigation access from mutation and automation access.

## Setup and configuration

The rad-security/mcp-server MCP server can be added as a remote HTTP MCP connection in supported clients. The documented configurations cover Claude Code, OpenAI Codex CLI, Cursor, VS Code with GitHub Copilot, Gemini CLI, Cline, Windsurf, Claude.ai, and Claude Desktop. Field names differ by client, but the endpoint and bearer credential remain the same.

Create a RAD Security API access key in the RAD Security console and provide it as `Bearer <access_key_id>:<secret_key>:<account_id>`. Codex can alternatively read the combined credential from an environment variable named `RAD_MCP_TOKEN`. The endpoint can also be tested with MCP Inspector or a JSON-RPC `curl` request using Streamable HTTP headers.

## Tools and capabilities

Available tool groups support:

- Listing and inspecting clusters, containers, images, and Kubernetes resources.
- Finding image vulnerabilities, top vulnerable images, SBOM data, and CVE dispositions.
- Reviewing process trees, runtime baselines, and container process analysis.
- Searching security findings, inbox items, and Kubernetes audit records for pod shell activity.
- Listing workflows, schedules, and runs; inspecting workflow definitions and runs; and running, creating, or updating workflows.

## Limitations and notes

All tools require authentication and a RAD Security account. The server operates against the account represented by the supplied credential. Write-capable tools can change security finding status, CVE reporting dispositions, inbox classification, or workflow configuration, so a read-only header is appropriate for agents that should only investigate.

Custom workflow definitions must be supplied as YAML text rather than a file path. The definition is validated before deployment; a failed validation does not deploy the workflow. Updates are limited to automations created through the custom-workflow creation tool. The hosted endpoint exposes the full toolkit by default, so explicit scoping is needed when an agent should see fewer tools.

_Full upstream README: https://allmcps.com/mcp/rad-security-mcp-server/readme_