Panther MCP Server – AI Security Operations

Panther's MCP server connects compatible AI clients to the Panther security operations platform. It lets agents assist with detection engineering, alert triage, log investigation, metrics review, and Panther access-context queries while using the permissions of the configured Panther API token.

#secops#detections#alerts

Overview

Panther MCP Server is an official open-source Model Context Protocol server from Panther Labs. It connects AI clients such as Claude Desktop, Claude Code, Cursor, and other MCP-compatible tools to a Panther instance so security teams can work with alerts, detections, log data, schemas, metrics, and access information from a conversational interface.

What the MCP server enables

The server exposes Panther-specific tools for common security operations workflows. An AI agent can list and inspect alerts, add comments, update alert status or assignee, retrieve alert events, and start or read AI alert triage summaries. For investigation work, it can run SQL queries against Panther's data lake, list databases and tables, fetch table schemas, inspect log sources, and summarize alert event patterns.

For detection engineering, the server provides tools to list, inspect, and disable detections across rules, scheduled rules, simple rules, and policies. It can also retrieve global helpers, data models, log type schemas, scheduled queries, alert metrics, ingestion metrics, Panther users, roles, and the current API token's permissions. The exact actions available depend on the Panther API token's granted permissions.

When to use it

Use Panther MCP Server when analysts, detection engineers, or security teams want AI assistance inside their development or investigation workflow without manually switching between an IDE, chat client, and Panther. Practical uses include triaging a high-severity alert, querying CloudTrail or other ingested log tables, reviewing detection code and tests, finding disabled policies, checking alert destination configuration, or exploring log schemas before writing SQL.

Connection and authentication

Panther supports local stdio usage and a persistent Streamable HTTP mode. The recommended local installation uses the official container image ghcr.io/panther-labs/mcp-panther; Python users can also run the package with uvx mcp-panther. Authentication is configured with PANTHER_INSTANCE_URL and PANTHER_API_TOKEN. The instance URL must include https://. Panther recommends starting with a read-only API token and storing the token securely.

Key considerations

The MCP server can access sensitive security data and, with sufficient token permissions, modify alerts or disable detections. Use least-privilege Panther API tokens, rotate credentials regularly, apply IP allow-lists where appropriate, and run the server in a constrained environment such as Docker. The HTTP mode is documented for local http://localhost:<port>/mcp usage and should be protected carefully before exposing it beyond a trusted network.

Frequently Asked Questions

When should an AI agent use Panther MCP Server?

Use it when an AI agent needs to assist with Panther security operations workflows such as alert triage, detection review, log investigation, data lake querying, schema exploration, or reviewing Panther metrics from an MCP-compatible client.

What Panther capabilities does the MCP server expose?

The server exposes tools for Panther alerts, data lake queries, scheduled queries, log sources, detections, global helpers, data models, log type schemas, metrics, users, roles, and current-token permissions. Available actions are limited by the configured Panther API token.

Can the server make changes in Panther?

Yes, if the API token has the necessary permissions. Documented tools include adding alert comments, updating alert status or assignee, bulk-updating alerts, and disabling detections. Teams should start with a read-only token unless write actions are intentionally required.

How is Panther MCP Server authenticated?

Authentication uses a Panther API token provided through `PANTHER_API_TOKEN` and a Panther instance URL provided through `PANTHER_INSTANCE_URL`. Panther's documentation recommends storing the generated token securely, applying least privilege, rotating credentials, and using IP allow-lists where appropriate.

Which transport should be used for Panther MCP Server?

Use stdio for local MCP clients such as Claude Desktop, Claude Code, Cursor, or Goose. Use Streamable HTTP when running Panther MCP Server as a persistent local or internal web service for multiple clients or long-running deployments, and protect the HTTP endpoint carefully.