Aimock

aimock, from the CopilotKit team, is a deterministic mock server for testing AI applications. It stands in for the services an AI app talks to, returning fixed responses defined in JSON fixtures, so tests run the same way every time in CI without API keys, network flakiness or model costs. It is built for developers writing automated tests for LLM applications and agents. Fixtures match on request content such as the user message and can simulate streaming chunk sizes and latency. Beyond LLM endpoints for chat completions, messages and embeddings across OpenAI, Claude and Gemini style APIs, one configuration file can mount mocks for MCP tools and resources, A2A agents, AG UI, vector databases such as Pinecone, search and rerank, and moderation, all on a single port. Record and replay mode proxies unmatched requests to the real provider, saves the responses to disk as editable JSON fixtures, and replays them on later identical requests; a proxy only mode forwards unmatched calls without writing anything. It runs through npx, as a Docker image, or as an npm library. The site publishes no pricing and no paid tiers. The tool is distributed as a public npm package and container image, so there is no license fee to evaluate. The vendor positions it as a suite that grows with the stack: teams can start with LLM mocking alone and add MCP, A2A and vector mocks later, and the LLM server itself is started with the llmock command.