AI Agents Framework — One YAML = Any Agent

Overview

AI Agents Framework is an open-source Rust framework built around the idea that one YAML file equals any agent. Developers write a single YAML specification to define and run AI agents that have identity, memory, behavioral structure, and tool-calling ability.

Key Capabilities

• YAML-driven definition: Declare agents through a single YAML specification, with --agent used only when the suite does not declare an agent, and --dry-config-check to validate the suite and referenced agent without selecting scenarios, constructing providers, or writing reports.
• Tool calls: Agents can make tool calls during a turn; the AgentResponse struct contains the final text content and an optional list of tool calls.
• Streaming responses: Supports AgentStreamEvent streams including Chunk and Final events for observing output in real time.
• Memory and sessions: Supports saving and listing sessions, with {{ context.<key> }} values from the context manager.
• Observability: Provides ObservabilityConfig and ObservabilityManager with callbacks for tool start and error events.
• CLI execution: Run agents from the command line with ai-agents-cli run agent, with RUST_LOG controlling log levels.
• Model integration: Named fields such as num_ctx, num_gpu, and keep_alive are merged into the Ollama request body.

Audience

Built for open-source contributors, AI agent builders, and Rust developers. The learning curve is medium, the monetization model is free, and it offers a public API and GitHub repository.