Thehive MCP Server
TheHive security platform interaction for AI agents via MCP-compliant API bridges.
Data last scanned today · Reviewed today
Overview
TheHiveMCP serves as a translation layer between AI models and TheHive security platform. Built in Go, it maps natural language queries into TheHive API calls, enabling agents to search for alerts, cases, tasks, and observables while supporting full CRUD operations on security entities. The server implements the MCP 1.0 specification and supports both Stdio and HTTP/SSE transport modes. It facilitates advanced security workflows, including the execution of Cortex analyzers, responders, and the promotion of alerts to cases directly through an AI assistant.
Our verdict
This server carries a high-risk score of 30/100, primarily because it is a community-maintained project rather than an official vendor release. With no external editorial review and a bus factor limited to a single contributor, users should exercise caution before deploying this in environments handling sensitive security data.
- Zero direct dependencies reduces supply chain attack surface
- Implements full MCP 1.0 protocol specification
- Supports both Stdio and HTTP transport modes
- Community-maintained rather than official vendor release
- Limited to one contributor presenting bus-factor risk
- Security-sensitive authentication methods currently unreviewed
Tools
| Tool | Description | Risk |
|---|---|---|
| search-entities | Search for entities with a structured filter built from TheHive's query DSL. | low |
| manage-entities | Create, update, delete entities, add comments, promote alerts to cases, merge cases/alerts/observables. | high |
| execute-automation | Run Cortex analyzers and responders, check job and action status. | high |
| get-resource | Access schemas, docs, and metadata through hierarchical browsing. | low |
Compatibility
| Client | Local | Docker | Remote | Read-only |
|---|---|---|---|---|
| ChatGPT | ||||
| Claude Desktop | ||||
| Cursor | ||||
| VS Code | ||||
| Windsurf |
Frequently asked questions
›Does this server handle authentication for incoming HTTP requests?
No, the server serves plain HTTP and does not perform authentication itself. It must be deployed behind a TLS-terminating, authenticating reverse proxy.
›What transport modes are supported for connecting the MCP server?
The server supports both stdio for local, single-user operations and HTTP with SSE support for scalable deployments.
›Which AI models are recommended for use with this server?
Use only the models specifically listed as recommended, such as claude-sonnet-4.6, gpt-5.5, or gemini-3.5-flash, as these have been verified for accuracy and prompt-injection resilience.
›How does the server mitigate the risk of prompt injection from security data?
The server wraps user-generated fields, such as titles and descriptions, in UNTRUSTED_DATA boundary tags and instructs the AI model not to follow instructions found within them.
›What is the default security configuration for tool execution?
The server defaults to a read-only configuration, which serves as an enforced backstop for security.
›How can I restrict which TheHive instances the server can interact with?
You can restrict the target TheHive URL by configuring the THEHIVE_URL or THEHIVE_URL_ALLOWLIST environment variables.
Alternatives
Changelog
Badge
Maintain this server? Add the live badge to your README.