MCP Victorialogs

What mcp-victorialogs MCP server does

The mcp-victorialogs MCP server gives an MCP client read-only access to a VictoriaLogs deployment. It works with single-node and cluster installations through the VictoriaLogs HTTP API, and also includes a locally embedded copy of VictoriaLogs documentation that can be searched without network access.

Its supported use cases center on log investigation and observability. An agent can query log records, inspect the available streams, enumerate fields and field values, retrieve instance parameters, and obtain query statistics in a metrics-oriented form. The available functionality corresponds broadly to the read-only operations exposed by the VictoriaLogs web UI.

How it works

The server receives requests from an MCP client and sends the relevant API requests to the configured VictoriaLogs endpoint. The target URL is supplied through configuration rather than selected by the client for each installation. Optional bearer authentication and custom request headers allow the server to reach protected or specially configured instances.

Two operating modes are described: stdio for local client integrations and Streamable HTTP for network-accessible use. Streamable HTTP mode also provides a root-endpoint UI with setup instructions and tool inspection. A public Streamable HTTP instance is available for trying the server against the VictoriaLogs playground, although it is intended for MCP clients rather than direct browser use.

Setup and configuration

Install the binary with Go, download a release artifact, run the published container, build from source, or use the available Helm chart. Building from source requires Go 1.26 or newer. A VictoriaLogs instance is required regardless of the installation method.

Set VL_INSTANCE_ENTRYPOINT to the VictoriaLogs URL. Authentication can use VL_INSTANCE_BEARER_TOKEN, while VL_INSTANCE_HEADERS accepts comma-separated key-value pairs for custom HTTP headers. In HTTP or SSE modes, MCP_PASSTHROUGH_HEADERS can forward selected headers from incoming MCP requests and takes precedence over matching configured headers. VL_DEFAULT_TENANT_ID supplies a fallback tenant in AccountID:ProjectID or AccountID form, defaulting to 0:0. Select the transport with MCP_SERVER_MODE; stdio is the default.

Tools and capabilities

The mcp-victorialogs MCP server supports capabilities for:

• Querying and exploring VictoriaLogs records
• Listing log streams
• Listing fields and values
• Reading VictoriaLogs instance parameters
• Returning log query statistics as metrics
• Searching embedded VictoriaLogs documentation
• Inspecting setup instructions and available tools through the HTTP root endpoint

The server is intended for read-only operations. Prompts can combine log queries with documentation searches, depending on the MCP client and model used.

Limitations and notes

This server does not provide write operations in the described API surface; its documented VictoriaLogs functions are read-only. Access still depends on a reachable VictoriaLogs instance and any required credentials or headers. The quality of generated investigations depends on the MCP client and model handling the returned tools and responses.

The mcp-victorialogs MCP server can be used with the public playground configuration for evaluation, but production deployments should point VL_INSTANCE_ENTRYPOINT at the intended VictoriaLogs service. The README specifically demonstrates Claude Desktop and does not document client-specific setup for other clients.