Codex App Server adapter for agent-kit: long-lived multi-turn thread client, one-shot structured-output client, and a surface-agnostic chat controller, all normalizing into the agent-core neutral schema.

A self-contained AI chat widget that drops into any web page with a single script tag. Powered by an agentic API backend, it supports streaming responses, multi-turn tool use, custom tool rendering, file uploads, and full UI customization — all isolated i

Agent execution layer: pluggable `ModelProvider` abstraction (`createProvider`, supports `anthropic` and `openai-compatible` kinds covering OpenAI, DeepSeek, Kimi, Together, Groq, Ollama), multi-turn tool-calling loop (`runAgent`), spec compliance checker

MCP server for Google AI Studio (Gemini) image generation with support for text-to-image, image editing, and multi-turn conversational image creation

Stratus Agent kernel: session, provider, tool, memory, and orchestration contracts with a multi-turn agent loop

Community TypeScript SDK + CLI for LightChain AI: end-to-end encrypted on-chain inference (with optional web search), multi-turn chat, batch + agent runners, read-only network analytics, an Ethereum bridge and LCAI Governor wrapper, and a full worker-oper

A local MCP server that gives unimodal LLMs multimodal vision — image/video recognition with multi-turn iterative analysis.

MCP server: generate and edit images via Gemini with Nano Banana multi-turn conversation support

N8n node for branch flow control and multi-turn conversations with context memory

MCP server bridging Claude Code (and other MCP clients) to a local LM Studio instance. Exposes model lifecycle (list/load/unload/download), single-turn inference with reasoning and stats, multi-turn chat with SQLite-persisted history, and embeddings.

SpecTalk Agent — multi-turn spec-writing agent for non-developers

Testing framework for OpenClaw bots. Spin up real agents, load skills, drive multi-turn prompts, and capture results.

ACP (Agent Client Protocol) stdio bridge for headless coding CLIs without a native or ecosystem ACP adapter — currently agy (Antigravity). Used by wicked-core to drive governed multi-turn sessions.

Generate text and chat completions using Cohere's Command family of large language models, with support for multi-turn conversations, retrieval augmented generation (RAG), tool use, reasoning, vision, structured output, and translation. Create text and im

LLM-powered manifest generation for Averos. Translates natural language application intent into validated JSON IR via cloud and local LLM providers — with automatic retry, validation feedback injection, multi-turn conversation, and seamless handoff to the

MCP evidence layer bridging Claude Code and OpenAI Codex — run-scoped git diff, sensitive content protection, multi-turn session management

Adversarial persona testing for AI agents — catch multi-turn failures before users do.

A2A (Agent-to-Agent) protocol wrapper for the GitHub Copilot SDK. Drop a JSON config file in, get a fully-spec-compliant A2A server out. Supports any agent persona, MCP tool servers, streaming, context building, and multi-turn conversations.

Middleware to destroy caching

Turn a function into an `http.Agent` instance