Agent Nexus

Agent Nexus is an open-source AI agent orchestration platform that lets developers build, deploy, and monitor multi-agent systems on their own infrastructure. It is model-agnostic, working with Anthropic Claude, OpenAI GPT, Google Gemini, and local Ollama models, so teams keep full data sovereignty and avoid vendor lock-in. The agent builder lets you configure each agent's system prompt, model, temperature, and max tokens, while a live playground streams the agent's reasoning in real time through an SSE trace panel and saves every session to a replayable conversation history. For more complex automations, a visual canvas editor built on React Flow provides drag-and-drop workflow construction, and webhook triggers expose persistent inbound HTTP endpoints that fire an agent or workflow run on demand. Governance features include risk-based approval gates that pause any tool marked requires_approval until a human approves, full run traces logging memory retrieval, context retrieval, model calls, and tool calls, isolated multi-workspace support, and role-based access control across owner, admin, member, and viewer roles. It is aimed at engineering teams, enterprises needing self-hosting and observability, and anyone building production multi-agent applications. The project is MIT-licensed and distributed for self-hosting.

Building multi-agent AI systems, orchestrating agent workflows visually, self-hosting AI agents for data sovereignty, running model-agnostic agents across Claude/GPT/Gemini/Ollama, adding human approval gates to agent runs, monitoring agent runs with full trace logs