Skip to content

SineFrame M3v0.2.16

Main Navigation

Test MCP servers and the agents that use them

Write your first MCP test

Install and update M3

Test your own server

Test a stdio server

Test a Streamable HTTP server

Test resources and prompts

Test tool errors and input schemas

Test state across tool calls

Test an agent’s tool use

Choose an agent harness

Pin a managed harness runtime

Run an ACP agent

Test a multi-turn agent session

Run agent cases as a matrix

Open saved test results

Read a finalized trace

Save and reopen executions

Compare a run with a baseline

Assert the evidence that matters

Write a custom evaluator

Evaluate a response with an LLM judge

Aggregate saved evaluations

Plan answers to elicitation requests

Submit input to a paused execution

Run M3 tests in CI

Run M3 in GitHub Actions

Publish or retry a CI run

Manage access tokens

Use the async SDK

Use mocks and replay

Snapshot stable values

Run M3 without pytest

Examples by task

Direct tests and agent tests

Runtime and connection lifecycle

Runs, suites, cases, and executions

Evidence and traces

Test, execution, and evaluation outcomes

Isolation and repeatability

Python SDK reference

Core execution API

Public value types

Observability models

Elicitation and managed-input API

Testing utilities

Python API inventory

pytest integration

Configuration and credentials

Compatibility and current boundaries

Troubleshooting

Installation and project Python

Server startup and connection failures

Agent readiness, approvals, and timeouts

Missing traces or saved results

CI credentials and uploads

For agents LLM index

Appearance

On this page

Migration

M3 v0.2.16 · commit 796f6943274f075d973f267794444b91ba208010.

This page has moved

The documentation is at /guides/evaluations/assertions.

Pager

Next pageTest MCP servers and the agents that use them