MCP Fusion

Command-Line Runner ​

The FusionTester is runner-agnostic — it returns plain JavaScript objects. You can use any test runner you want: Vitest, Jest, Mocha, or Node's native node:test. This page documents the most common CLI patterns using Vitest (recommended).

Running Tests ​

Run All Tests ​

bash
npx vitest run

Output:

 ✓ tests/firewall/user.firewall.test.ts (3 tests) 4ms
 ✓ tests/guards/user.guard.test.ts (4 tests) 3ms
 ✓ tests/rules/user.rules.test.ts (3 tests) 2ms
 ✓ tests/blocks/analytics.blocks.test.ts (3 tests) 2ms
 ✓ tests/guards/user.oom.test.ts (5 tests) 2ms

 Test Files  5 passed (5)
      Tests  18 passed (18)
   Duration  520ms

Run with Verbose Output ​

bash
npx vitest run --reporter=verbose

Shows every individual test with timing:

 ✓ tests/firewall/user.firewall.test.ts (3 tests) 4ms
   ✓ User Egress Firewall > strips passwordHash from response 2ms
   ✓ User Egress Firewall > strips tenantId (multi-tenant isolation) 1ms
   ✓ User Egress Firewall > preserves declared fields accurately 1ms

Selecting Tests ​

By Directory (Governance Concern) ​

bash
# Egress Firewall audits only
npx vitest run tests/firewall/

# Middleware & OOM Guard audits only
npx vitest run tests/guards/

# System Rules audits only
npx vitest run tests/rules/

# UI Blocks audits only
npx vitest run tests/blocks/

By File (Specific Entity) ​

bash
# User firewall tests only
npx vitest run tests/firewall/user.firewall.test.ts

# Order guard tests only
npx vitest run tests/guards/order.guard.test.ts

Multiple Directories/Files ​

bash
# Firewall + Rules (skip guards and blocks)
npx vitest run tests/firewall/ tests/rules/

# Specific files
npx vitest run tests/firewall/user.firewall.test.ts tests/rules/user.rules.test.ts

Filtering by Test Name ​

Use -t (or --testNamePattern) to filter by test description:

bash
# Only tests containing "passwordHash"
npx vitest run -t "passwordHash"

# Only tests containing "GUEST"
npx vitest run -t "GUEST"

# Only tests containing "OOM"
npx vitest run -t "OOM"

# Only truncation tests
npx vitest run -t "truncat"

# Only tests for ADMIN role
npx vitest run -t "ADMIN"

# Only email validation tests
npx vitest run -t "email"

# Only "strips" tests (egress firewall)
npx vitest run -t "strips"

Combining Directory + Name Filter ​

bash
# Only user firewall tests that mention "password"
npx vitest run tests/firewall/user.firewall.test.ts -t "password"

# Only guard tests for GUEST role
npx vitest run tests/guards/ -t "GUEST"

# Only OOM tests about boundaries
npx vitest run tests/guards/ -t "rejects"

Watch Mode ​

Re-runs tests automatically when source files change — essential during development:

bash
# Watch all tests
npx vitest watch

# Watch only firewall tests
npx vitest watch tests/firewall/

# Watch with name filter
npx vitest watch -t "passwordHash"

# Watch specific file
npx vitest watch tests/guards/user.guard.test.ts

Watch mode re-runs only the affected tests when you edit a source file:

RERUN  tests/firewall/user.firewall.test.ts x 1

 ✓ tests/firewall/user.firewall.test.ts (3 tests) 3ms
   ✓ User Egress Firewall > strips passwordHash from response 2ms
   ✓ User Egress Firewall > strips tenantId (multi-tenant isolation) 0ms
   ✓ User Egress Firewall > preserves declared fields accurately 0ms

 Test Files  1 passed (1)
      Tests  3 passed (3)
   Duration  180ms

 PASS  Waiting for file changes...
       press h to show help, press q to quit

Coverage ​

bash
# Run with coverage
npx vitest run --coverage

# Coverage for specific directory
npx vitest run --coverage tests/firewall/

# Coverage with specific reporter
npx vitest run --coverage --coverage.reporter=text --coverage.reporter=html

Example output:

 % Coverage Report
 --------------------------------|---------|----------|---------|---------|
 File                            | % Stmts | % Branch | % Funcs | % Lines |
 --------------------------------|---------|----------|---------|---------|
 src/views/user.presenter.ts     |   100   |   100    |   100   |   100   |
 src/views/order.presenter.ts    |    92   |    85    |   100   |    92   |
 src/agents/user.tool.ts         |    95   |    90    |   100   |    95   |
 src/agents/order.tool.ts        |    88   |    80    |   100   |    88   |
 src/models/user.schema.ts       |   100   |   100    |   100   |   100   |
 src/models/order.schema.ts      |   100   |   100    |   100   |   100   |
 --------------------------------|---------|----------|---------|---------|
 All files                       |    96   |    92    |   100   |    96   |
 --------------------------------|---------|----------|---------|---------|

Output Formats ​

Default (dots) ​

bash
npx vitest run

Verbose (every test listed) ​

bash
npx vitest run --reporter=verbose

JSON (machine-readable) ​

bash
npx vitest run --reporter=json --outputFile=results.json

JUnit XML (CI/CD integration) ​

bash
npx vitest run --reporter=junit --outputFile=results.xml

Multiple reporters simultaneously ​

bash
npx vitest run --reporter=verbose --reporter=junit --outputFile.junit=results.xml

Exit Codes ​

CodeMeaning
0All tests passed
1At least one test failed
2Configuration error

Stopping Early ​

bash
# Stop on first failure
npx vitest run --bail 1

# Stop after 5 failures
npx vitest run --bail 5

Test Execution Order ​

bash
# Randomize execution order
npx vitest run --sequence.shuffle

# Run in parallel (default)
npx vitest run --pool=threads

# Run sequentially  
npx vitest run --pool=forks --poolOptions.forks.singleFork

Jest Equivalent Commands ​

If your project uses Jest instead of Vitest:

bash
# Run all tests
npx jest

# Run by directory
npx jest tests/firewall/

# Filter by name
npx jest -t "passwordHash"

# Watch mode
npx jest --watch

# Coverage
npx jest --coverage

# Verbose
npx jest --verbose

# Bail on first failure
npx jest --bail

Node.js Native Test Runner ​

If you prefer zero dependencies:

bash
# Run all tests
node --test tests/

# Run specific file
node --test tests/firewall/user.firewall.test.ts

# Filter by name
node --test --test-name-pattern="passwordHash" tests/

No API Keys Required

Notice what's missing from every command above: OPENAI_API_KEY, ANTHROPIC_API_KEY, CLAUDE_API_KEY. The FusionTester runs 100% in RAM. No LLM calls. No token costs. No rate limits. No flaky tests from API outages.