Back to Ecosystem Pulse
ProtocolExperimentalA2A
ai-mocks
by mokksy
Kotlin mock server for LLM and A2A endpoints with real streaming and SSE
Kotlin
Updated Jul 24, 2026
Share:
Overview
Emulates LLM and agent-to-agent endpoints locally to enable reliable integration and contract testing. Runs a Kotlin-based mock server with real streaming and SSE behavior so clients see the same timing, chunking, and error modes as production LLMs. Includes mocks for OpenAI, Anthropic, Gemini, Ollama and A2A-style services so you can test client logic, streaming handlers, and failure cases without network dependencies.
Why It Matters
As multi-agent systems rely on live LLM services and A2A endpoints, small differences in streaming, retries, or error shapes cause brittle systems and hidden failure modes. Having faithful local mocks makes A2A evaluation A2A evaluation and continuous testing practical: you can reproduce race conditions, partial streams, and protocol edge cases consistently. That visibility is essential for building agent track records and automated evaluation pipelines before deploying to production.
Ideal For
Developers and QA teams building or testing agents and client SDKs who need realistic local LLM/A2A behavior for integration and regression tests, including regression tests.
Applications
- Simulate streaming responses and SSE to reproduce agent interaction timing or chunking bugs
- Run offline integration tests for agent clients and SDKs that expect OpenAI/Anthropic/Gemini APIs
- Inject deterministic failure modes (timeouts, partial streams, protocol errors) for continuous agent evaluation
- Validate A2A message handling and retry logic in multi-agent workflows without calling real endpoints
Works With
openaianthropic
Topics
a2a-protocolagent-to-agentkotlinktorktor-serverlangchain4jllmmockmock-serveropenai-api+3 more
Similar Tools
agent-playgroundautogen
Keywords
a2a evaluationmulti-agent trustagent-to-agent evaluationstreaming mocks