aimock

aimock provides deterministic mock infrastructure for AI applications, enabling developers to simulate and test interactions with a wide range of AI-related services. It addresses the challenge of reliably testing AI apps that depend on external APIs, such as large language models (LLMs), multi-component protocols (MCP), agent-to-agent (A2A) systems, vector databases, and multimedia endpoints. By recording and replaying API interactions as fixtures, aimock ensures consistent and reproducible test results without requiring live API keys or risking flaky tests.

The platform supports mocking of major LLM providers, including OpenAI, Claude, Gemini, Bedrock, Azure, Vertex AI, Ollama, Cohere, and ElevenLabs, with full streaming and embeddings support. It also enables simulation of MCP tools and resources, A2A agent interactions with message routing and server-sent events (SSE) streaming, and AG-UI protocol event streams for frontend testing. For vector databases, aimock is compatible with Pinecone, Qdrant, and ChromaDB, allowing developers to mock similarity search, upserts, and index operations. Additionally, it covers multimedia APIs for image generation and editing, text-to-speech, audio transcription and translation, non-speech audio generation, and video generation, all configurable via fixtures.

aimock offers robust features for record and replay workflows: it can proxy unmatched requests to real APIs, automatically save responses as editable JSON fixtures, and replay them deterministically in continuous integration environments. The tool supports chaos testing by allowing users to simulate dropped, malformed, or disconnected responses with configurable probabilities to verify app resilience. It also includes drift detection, automatically monitoring real API endpoints in daily CI runs, validating response schemas, and updating fixtures to keep tests current as provider APIs evolve.

js library. It integrates with testing frameworks like Vitest and Jest, providing plugins for server lifecycle management, environment variable patching, and match count resets. Configuration is managed through a single JSON file that specifies fixtures and providers for each service type. The tool is designed to grow with the complexity of an AI application's stack, supporting a variety of protocols and services as needed.