There are a few tools in the space, but they each focus on parts:
- MCP Inspector (Anthropic's official tool) — interactive debugging, not CI-oriented
- MCP-Scan (Invariant Labs, now Snyk) — security scanning, focused on tool poisoning detection
- Promptfoo — LLM red teaming tool that added MCP support recently
- MCP Protocol Validator — spec compliance checking
MCPSpec tries to be the one tool that covers the full workflow: record, replay, mock, security audit, quality scoring, and CI setup. None of the above do recording/replay or mock generation.
As for standards — there aren't any yet. MCP itself moved under the Linux Foundation's Agentic AI Foundation in December 2025, and NIST launched an AI Agent Standards Initiative last month. But no conformance or testing standard exists. It's still early.
- MCP Inspector (Anthropic's official tool) — interactive debugging, not CI-oriented - MCP-Scan (Invariant Labs, now Snyk) — security scanning, focused on tool poisoning detection - Promptfoo — LLM red teaming tool that added MCP support recently - MCP Protocol Validator — spec compliance checking
MCPSpec tries to be the one tool that covers the full workflow: record, replay, mock, security audit, quality scoring, and CI setup. None of the above do recording/replay or mock generation.
As for standards — there aren't any yet. MCP itself moved under the Linux Foundation's Agentic AI Foundation in December 2025, and NIST launched an AI Agent Standards Initiative last month. But no conformance or testing standard exists. It's still early.