Seristack

What seristack does

The seristack MCP server turns shell-based workflows described in YAML into callable MCP tools. Each stack can contain one or more commands, a working directory, variables, execution settings, dependencies, and optional output processing. Selected stacks can be exposed to AI agents and IDE integrations while the same configuration remains usable from the command line or HTTP server.

This approach fits operational tasks that are currently spread across shell scripts, runbooks, or CI steps. A stack can represent a health check, deployment action, diagnostic routine, or other command sequence, provided the required commands are available in the execution environment.

How it works

A YAML file defines named stacks and their behavior. Seristack can run commands sequentially or concurrently, repeat a stack a configured number of times, and continue or stop when a command fails. One stack may depend on another, and dependent stacks can consume shared results. Variable definitions support required values and allowed-value restrictions, while command templates can reference variables and execution-count data.

The MCP mode is started with the mcp command. The supplied example uses streamable HTTP transport, allowing an MCP-capable agent or IDE to connect to the running process. The same stack definitions can also be triggered directly with the CLI or served through HTTP routes.

Setup and configuration

Install the binary through Homebrew on macOS or Linux, the provided Linux or Windows installer, or a release archive. Create a YAML configuration containing the stacks to expose, then start MCP mode with a command such as seristack mcp -t streamableHTTP. The excerpt does not specify a configuration-file flag for that command, so consult the project’s configuration reference and command help for the applicable runtime options.

For HTTP or remote MCP deployments, the documented pattern binds Seristack to localhost and places a reverse proxy or gateway in front of it. That proxy can provide TLS, authentication, rate limiting, and trusted identity headers. Do not expose command execution directly to the public internet.

Tools and capabilities

Configured stacks can provide:

• Shell execution through mvdan shell, Bash, sh, or PowerShell
• Sequential or parallel execution
• Repetition counts and per-stack timeouts
• Dependencies and result sharing
• Working-directory selection
• Variable validation, substitution, and allow/deny rules
• HTTP endpoint exposure
• Per-stack authorization based on identity headers
• Structured JSON audit logging

The seristack MCP server does not define a fixed catalog of built-in domain tools in the provided material. Its available MCP operations come from the stacks configured by the operator.

Limitations and notes

Because stacks execute shell commands, the host environment determines which programs, scripts, permissions, and shells are available. Configuration and authorization therefore require careful review. Identity-header access rules should only be trusted when requests come through the authenticated proxy that sets them; untrusted clients must not be allowed to supply those headers directly.

The provided material documents streamable HTTP startup but does not name specific MCP clients or provide client-specific setup instructions. It also does not state a license name beyond linking to a repository license file.