Snouty: Antithesis test runs from an agent
snouty
Official Antithesis CLI with an MCP server: an agent launches runs, reads properties, logs and events and checks docker-compose
- 33 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- An agent launches and cancels runs in the Antithesis account and pushes build images to the registry
- The exec command runs a bash script inside a run in progress
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/snoutyDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- An agent launches and cancels runs in the Antithesis account and pushes build images to the registry
- The exec command runs a bash script inside a run in progress
This is third-party code. Review the repository files before installing.
What it does
Snouty is a command line utility for the Antithesis API, a platform for deterministic testing of distributed systems. The snouty mcp command starts an MCP server over stdio and gives an agent most of the utility's commands as tools. Through them the agent launches a run and pushes images, lists runs with filters, shows the status of one run, waits for it to be ready or finish and cancels one in progress. It reads property results, meaning checks and assertions, streams build logs and logs along one branch, searches events and runs a query in the event-set language. The agent also starts a debugging session, validates a docker-compose.yaml setup locally with validate, searches the Antithesis documentation and runs a bash script at a chosen moment of a run in progress. Many commands support JSON output.
Who it is for. Engineers who test distributed systems on Antithesis and want Claude Code or Codex to launch runs and dig into failed properties on its own.
Good fit when
- After a service change you launch a run and wait until it is ready to inspect
- A property failed and you want to find the event in the logs of the branch where it happened
- Before a cloud run you check locally that docker-compose.yaml builds and reports readiness
- A run in progress has to be stopped because a wrong build went into it
- You look up in the Antithesis docs how to describe a property without leaving the agent
Not a fit when
- You do not use Antithesis: the utility works only with its API
- You need a server with no prior sign-in: run snouty login once first
- The agent must not start paid runs on its own: switch off the launch tools or do not connect the server
- There is no container engine on the machine: commands with docker-compose.yaml do not work without it
Example request
Launch an Antithesis run for the current build, wait until it is ready and show which properties failed and in which event.MCP
- Transport
- stdio
- Authentication
- not required
Limitations
Run snouty login before connecting: the server takes the utility's credentials and settings. The MCP server has no authentication of its own and talks to the client only over stdio. Running and validating need Docker Compose version 2 and Docker or Podman, and with both engines Podman is preferred.
The server sets no time or output limit, the client does: when a call is cancelled, the command process is killed. Up to eight calls run at once, and the --max-concurrent-calls flag changes that number. The exec command needs tenant release 64.0 or newer.
How to disable. Remove the server with claude mcp remove snouty. Credentials from snouty login live in the utility's settings and can be reset by signing in again or deleting the config.
FAQ
How do I connect Snouty MCP to Claude Code?
Run snouty login, then claude mcp add snouty --scope user -- snouty mcp. The --scope user flag makes the server available in every project, and claude mcp list checks the connection.
How do I install snouty itself?
There are prebuilt binaries through an install script and cargo binstall, an npm package, a source build through cargo install and a package in nixpkgs. It updates with snouty update.
Does the MCP server need a token?
There is no separate token: the server uses credentials from snouty login. The server itself has no password and listens only on stdio, so it is reachable by the process that started it.
How many calls run at once?
Up to eight. The number is changed with the --max-concurrent-calls flag when starting snouty mcp. The server has no time limit, and when the client cancels a call the matching command process is killed.
How do I debug the server?
Set RUST_LOG=debug and add --verbose, for example claude mcp add snouty -e RUST_LOG=debug -- snouty --verbose mcp. The server log goes to stderr, and in Claude Code it shows through claude --debug and /mcp.
Discussion
No questions or reviews yet.
Related
MCP serversAntithesis Skills
antithesis-skills
Official Antithesis skills: the agent analyzes your system, picks reliability properties, builds a harness, launches tests and triages the results
A skill and MCP server for the Rhesis platform: the agent designs, runs and analyzes test suites for your own AI agent
An MCP server that checks whether the agent is really done: drives the real app and returns pass, fail or couldn't tell with the file and line
The CLI generates API tests from schemas and gives agents reproducible failures in responses and operation sequences.
The SmartBear server gives the agent BugSnag, Swagger, Zephyr, QMetry and PactFlow: errors, API specs, test cases and contracts
kogiQA MCP gives an agent a browser that finds page elements without selectors, for debugging web apps and writing e2e tests
README in short
The README lists snouty subcommands, requirements for Docker Compose and a container engine, install methods and utility settings. A separate section covers the MCP server, connecting to Claude Code and Codex, generic JSON and debugging.