Crate llmy_agent 

Building blocks for tool-using LLM agents.

llmy-agent provides the small, opinionated primitives that higher-level crates (such as llmy-harness) compose into a complete agent loop:

• tool — the Tool / ToolDyn traits and the tool::ToolBox registry, used to expose Rust functions to a model and to dispatch tool calls back to them.
• agent — the StepResult type that summarises a single turn of an agent loop (the model either stopped with a final assistant message or issued tool calls that need to be executed before the next step).

Defining a tool

Implementing Tool by hand is straightforward, but the llmy_agent_derive::tool attribute macro (re-exported here as tool, and also reachable through the umbrella crate as llmy::agent::tool) generates the boilerplate for you:

use llmy_agent::{LLMYError, tool};
use schemars::JsonSchema;
use serde::Deserialize;

#[derive(Deserialize, JsonSchema)]
struct AddArgs { a: i64, b: i64 }

#[derive(Clone, Debug)]
#[tool(
    description = "Add two integers",
    arguments = AddArgs,
    invoke = run,
)]
struct AddTool;

impl AddTool {
    async fn run(&self, args: AddArgs) -> Result<String, LLMYError> {
        Ok((args.a + args.b).to_string())
    }
}

Tools can then be collected into a tool::ToolBox, rendered as OpenAI tool descriptors with tool::ToolBox::openai_objects, and dispatched against incoming tool_calls with the various invoke_* methods.

Re-exports

For convenience the most commonly used items are re-exported at the crate root: StepResult, Tool, ToolDyn, LLMYError and the tool attribute macro.

Re-exports

pub use agent::StepResult;

pub use tool::Tool;

pub use tool::ToolDyn;

Modules

agent

Outcome of one step of an agent loop.

mcp

MCP server support — serve a tool::ToolBox over stdio or HTTP. MCP server support — expose a [ToolBox] as an MCP server over stdio or Streamable HTTP.

rcmp

Re-export of the rmcp crate for MCP protocol types.

tool

Tool definitions and the ToolBox registry used by agents.

Enums

LLMYError

Attribute Macros

tool