Persona

Persona is a developer platform for building, testing, and running LLM-powered agents from a single console. Core capabilities: • Organize agent work in Projects with configurable skills and instructions • Connect knowledge bases for retrieval-augmented generation (RAG) and MCP connectors • Add REST tools and choose any LLM provider • Test in the playground, run evals, and inspect traces • Ship versioned releases • Embed agents with the TypeScript SDK, React hooks, and chat UI Persona is designed so you can describe an agent in plain English and let the Architect draft its instructions and pick the right tools, or define the same agent in code with the SDK. You then connect your REST endpoints, MCP servers, and documents, while credentials stay encrypted inside your project. In the playground you can chat with the agent, approve or reject actions it wants to take, and replay any run with a different prompt or model to see how changes affect results. Every run is recorded as a trace showing each model call, each request to your own APIs, what was searched, and how long each step took and what it cost. Eval datasets help you measure quality before publishing a version, and the streaming runtime, built on AG-UI, delivers responses live to your users. Persona also keeps conversations, memory, and tool calls, so you do not need to build your own agent backend. Persona suits product teams, startups, and developers who want to add a support assistant, internal copilot, or customer-facing agent to an existing application. It is especially useful for teams that need reliable testing, version control, and observability without maintaining custom infrastructure.