Mcp Observatory
Regression-test your own MCP servers—capabilities, tool calls, and schema drift—before you ship or after you upgrade dependencies.
Overview
io.github.KryptosAI/mcp-observatory is an MCP server for the Ship phase that regression-tests MCP servers by checking capabilities, invoking tools, and detecting schema drift.
What is this MCP server?
- Regression testing purpose-built for MCP servers, not generic HTTP smoke tests
- Checks server capabilities, invokes tools, and detects schema drift
- npm package @kryptosai/mcp-observatory at version 0.3.2 with stdio transport
- Fits indie maintainers shipping multiple MCP tools to Claude Code and Cursor users
- Open source on github.com/KryptosAI/mcp-observatory
- npm package: @kryptosai/mcp-observatory version 0.3.2
- Transport: stdio
- Repository: github.com/KryptosAI/mcp-observatory
Community signal: 1 GitHub stars.
What problem does it solve?
Your MCP server looks fine until a schema change breaks agent tool calls and you only find out from users.
Who is it for?
Indie devs shipping or forking MCP servers who need automated contract tests in local or CI workflows.
Skip if: Builders who only consume third-party MCP and never maintain server definitions themselves.
What do I get? / Deliverables
You run repeatable MCP regression checks and get early warnings when capabilities or schemas drift.
- Regression run results for MCP capabilities and tool invocations
- Drift signals when tool schemas change versus a baseline
Recommended MCP Servers
Journey fit
Shipping agent products that depend on MCP means broken tools silently erode trust; automated checks belong in the ship gate. Testing subphase covers automated verification beyond unit tests—here specifically MCP contract and behavior regression.
How it compares
MCP-specific regression tester—not a general application monitor or penetration testing suite.
Common Questions / FAQ
Who is mcp-observatory for?
Developers who author or maintain MCP servers and want automated regression coverage before release.
When should I use mcp-observatory?
Before shipping MCP updates, after dependency bumps, or on a schedule when you operate custom MCP integrations.
How do I add mcp-observatory to my agent?
Install @kryptosai/mcp-observatory from npm (0.3.2), add it as a stdio MCP server in Claude Code or Cursor, and point it at the MCP servers you want to test per the repo README.