# About MCP Inspect and mcprobe

> MCP Inspect is a privately operated MCP contract testing product. Find its maintainer, official website, mcprobe CLI, support address and evidence boundaries.

MCP Inspect is a contract testing and monitoring product for engineers who
maintain MCP servers. It is maintained by
[Jovi De Croock](https://github.com/JoviDeCroock) and privately operated at
mcprobe.dev. It captures advertised contracts, compares versions with
deterministic rules, and keeps hosted history and scheduled checks.

## Official product locations

| Location                                                | Purpose                                                                    |
| ------------------------------------------------------- | -------------------------------------------------------------------------- |
| [mcprobe.dev](https://mcprobe.dev/)                     | Public product information, documentation and the free schema diff checker |
| [app.mcprobe.dev](https://app.mcprobe.dev/signup)       | Hosted account, contract capture, history and monitoring                   |
| [mcprobe on npm](https://www.npmjs.com/package/mcprobe) | The published local CLI; public recipes pin `mcprobe@0.1.1`                |
| [Agent skills](/docs/agents)                            | Public procedural instructions for setup and contract review               |

**MCP Inspect** is the product name; **mcprobe** is its npm CLI. The npm package
named `mcp-inspect` is a different project. The
[official MCP Inspector](https://modelcontextprotocol.io/docs/tools/inspector)
is a separate interactive debugging client from the MCP project. See
[MCP server testing](/docs/mcp-server-testing) for how the workflows fit together.

The source repository is private and self-hosting is not offered. The verified
CLI bundle provides local snapshot, diff and check without an account; public
instructions do not require access to private source. Instrumentation and the
standalone Action remain approved previews.

## What supports a compatibility finding?

Every gating severity comes from deterministic rules. The
[generated rule reference](/docs/rules) describes the classifier and migration
hint for each stable rule id. The [example lab](/examples) exposes illustrative
before/after snapshots and reproducible engine results. These examples are
product demonstrations, not customer contracts or testimonials.

Descriptions and instructions steer tool selection, so they belong to the
contract. Their classification is a heuristic: it does not measure a particular
model's behavior. A passing comparison does not prove that a handler works or
an agent completes its task. Integration tests and evaluations remain necessary.

Discovery never invokes tools. The inspector connects on demand, and scheduled
captures stay outside the request path. Optional telemetry records argument
paths, never values. Observed usage always needs a window and evidence limits;
hosted usage queries are currently pending. Read the
[privacy and evidence policy](/docs/privacy).

## Support and getting started

For product, pricing or contract-review questions, email
[support@mcprobe.dev](mailto:support@mcprobe.dev). A minimal before/after tool
definition or a stable rule id helps identify a contract question.

[Try the free checker](/mcp-schema-diff), read [Free and Team pricing](/docs/pricing),
or [start your first hosted capture](/docs/getting-started).

Reviewed October 14, 2026. This is the operator's product documentation; findings
and examples describe implementation behavior, not independent endorsements.