LI.FI API Testing Suite (2026)

A comprehensive API test suite for LI.FI endpoints, built with Playwright + TypeScript and Postman/Newman, with MCP server parity tests and AI-assisted scenario generation.

---

What's in this repo

Layer	Tool	Purpose
Playwright + TypeScript	playwright/tests/	Automated regression suite, CI/CD
Postman collections	postman/collections/	Manual/exploratory API validation
MCP parity tests	playwright/tests/mcp-parity.spec.ts	Validates LI.FI MCP server against REST API
AI scenario generator	mcp/mcp-test-scenarios.ts	Claude generates edge case test data
GitHub Actions	.github/workflows/	CI on PR + daily scheduled health monitor

---

Endpoints covered

Endpoint	Tests	Notes
GET /tokens	tokens.spec.ts	Cached — see caching strategy below
GET /chains	chains.spec.ts	Cached
GET /tools	tools.spec.ts	Cached
GET /connections	connections.spec.ts	Not cached — param-dependent
GET /quote	quote.spec.ts	Never cached — real-time routing
GET /quote (Composer)	composer.spec.ts	Composer deposit flows
MCP tools	mcp-parity.spec.ts	get-chains, get-quote, get-tokens parity

---

Caching strategy

The original Postman suite fetched /tokens on every test, resulting in repeated heavy payloads (~100k tokens across 60+ chains). This suite caches static-data endpoints for the duration of each test run:

/tokens  → cached (data changes infrequently, large payload)
/chains  → cached (chain list is stable)
/tools   → cached (bridge/DEX list is stable)
/quote   → NEVER cached (real-time routing, must be fresh)
/routes  → NEVER cached (real-time routing, must be fresh)

The LiFiApiClient class in playwright/helpers/api-client.ts handles this automatically. The caching test in tokens.spec.ts explicitly validates that the second call is <10ms.

---

MCP Integration

LI.FI launched their own MCP server in March 2026 at https://mcp.li.quest/mcp. This wraps the REST API into tools for AI agents (Claude, Cursor, Windsurf, etc).

mcp-parity.spec.ts validates that the MCP server returns data consistent with the underlying REST API — catching any divergence between the two surfaces.

To connect LI.FI's MCP server to Claude Desktop, add to your config:

{
  "mcpServers": {
    "lifi": {
      "type": "http",
      "url": "https://mcp.li.quest/mcp"
    }
  }
}

---

AI-assisted test scenario generation

mcp/mcp-test-scenarios.ts uses an LLM (Claude by default) to generate diverse edge case test scenarios across multiple endpoints — /quote, /connections, /tokens, /chains, and /tools.

npm run scenarios

Output is saved to mcp/generated-scenarios.json. The Playwright test suite consumes these scenarios automatically via playwright/helpers/agentic-scenarios.ts → getAgentScenarios(endpoint):

• quote.spec.ts — loads /quote scenarios (positive + negative + edge cases)
• connections.spec.ts — loads /connections scenarios

Deterministic fallback: if the JSON file is missing or the LLM call fails, the tests skip the agentic block gracefully — the deterministic suite still runs. To regenerate scenarios, set ANTHROPIC_API_KEY in .env and run npm run scenarios.

Postman/Newman can run the same generated scenarios by first producing iteration data and then running the Agentic Scenarios folder:

npm run postman:data:agentic
npm run test:postman:agentic

This keeps AI-driven tests deterministic in CI by using checked-in JSON inputs.

Each generated scenario includes:

Field	Description
endpoint	Target API path, e.g. /quote
params	Query parameters passed to the endpoint
expectedStatus	Allowed HTTP status codes
expectedBehaviour	valid_route | no_route | error | schema_violation
notes	Why this scenario is an interesting edge case

Screen.Recording.2026-04-07.at.21.30.26.mov

---

Setup

npm install
npx playwright install --with-deps chromium

Copy the environment template and fill in your keys:

cp .env.example .env

Variable	Required	Purpose
LIFI_API_KEY	No	Increases rate limits (200 req/2h → 200 req/min)
ANTHROPIC_API_KEY	Only for npm run scenarios	Claude-assisted scenario generation

---

Running tests

For a fast demo flow, see docs/POSTMAN-RUNBOOK.md.

Postman UI quick start

1. Import postman/collections/lifi_full_suite.postman_collection.json.
2. Import postman/environments/production.postman_environment.json.
3. Select environment LI.FI Production.
4. Run folder Connections (/connections) in Postman Runner.
5. For Agentic Scenarios, run via Runner with data file postman/data/agentic-scenarios.postman_data.json.

# All Playwright tests
npm test

# Single suite
npm run test:tokens
npm run test:quote
npm run test:composer
npm run test:mcp

# Postman collections via Newman
npm run test:postman
npm run test:postman:quote
npm run test:postman:connections
npm run test:postman:chains
npm run test:postman:tools

# AI scenario generation
npm run scenarios
npm run postman:data:agentic
npm run test:postman:agentic

# Staging environment (placeholder URL by default)
node scripts/newman-run.js --env=staging

# Type-check without running tests
npx tsc --noEmit

# View HTML report
npm run report

---

CI/CD

GitHub Actions runs on:

• Every pull request to main
• Every push to main
• Daily at 08:00 UTC (live API health monitor)

Results are uploaded as artifacts and available in the Actions tab.

---

Project structure

lifi-api-test-suite/
├── playwright/
│   ├── tests/
│   │   ├── tokens.spec.ts
│   │   ├── chains.spec.ts
│   │   ├── tools.spec.ts
│   │   ├── connections.spec.ts
│   │   ├── quote.spec.ts
│   │   ├── composer.spec.ts
│   │   └── mcp-parity.spec.ts
│   └── helpers/
│       ├── api-client.ts       # Centralised client with caching + type definitions
│       ├── assertions.ts       # Shared custom assertions
│       ├── agentic-scenarios.ts # Loader for AI-generated scenario JSON
│       └── test-data.ts        # Chain IDs, token addresses, test wallets
├── postman/
│   ├── collections/
│   ├── environments/
│   └── data/
├── mcp/
│   ├── mcp-test-scenarios.ts   # AI-assisted scenario generator
│   └── generated-scenarios.json
├── scripts/
│   ├── newman-run.js           # Newman CLI runner
│   └── generate-postman-agentic-data.js
├── .github/workflows/
│   └── api-tests.yml
├── .vscode/
│   ├── settings.json
│   └── extensions.json
├── docs/
│   └── TEST-PLAN.md
├── AGENTS.md                   # Context for AI agents and reviewers
├── playwright.config.ts
├── .env.example
└── package.json