Context Engineering for developers
Build, test and analyse agents behaviour using marrakesh prompts
Works with
Vercel AI SDK
OpenAI (soon)
Anthropic (soon)
Langchain (soon)
Define prompts as typed TypeScript code
Build AI agents with full type safety. Define system prompts, tools, and outputs in clean, maintainable code.
Test your prompts with evals
Run evaluations directly from your CLI. Validate agent behavior with executable test cases before deployment.
Terminal
$npx @marrakesh/cli test prompts/weather-agent.ts
Running tests... Found 1 test file
[1/2]What's the weather in Paris?
Get current weather for a city({"city": "Paris"})
✔Passed (5.35s)
[2/2]What's the forecast for Tokyo?
Get current weather for a city({"city":"Tokyo"})
✔Passed (1.28s)
+All tests passed! 2/2 (100.0%)
Time: 6.63s
Track and analyze every execution
Real-time analytics dashboard for your AI prompts. Monitor performance metrics, system health, and cost analysis.