Unit 9 · Connect to external tools and data

Handle long-running MCP work

Modern servers may return a task from an ordinary tool call, which Ax can auto-await or expose for input, cancellation, and observation. Legacy task-draft APIs remain compatibility-only.

11 focused minutesNot started

Worked example

See the idea in context

let task = client.lock().unwrap()
    .call_tool("start_reindex", json!({"scope":"all"}))?;
let task_id = task["task"]["taskId"].as_str().unwrap().to_string();
target.wait_for = vec![json!({
    "kind":"mcp.task", "value":"taskKey",
    "metadata":{"taskId":task_id},
})];
runtime.register_target(target);
runtime.start()?;
runtime.publish(start_event, "tenant:demo", "authenticated")?;
let runtime = Arc::new(Mutex::new(runtime));
let mut source = AxMCPEventSource::new(
    client.clone(), runtime.clone(), "inventory",
    "tenant:demo", "authenticated", vec![],
);
source.start()?;
source.poll(); // Terminal task status resumes the owner.

Run itIn your own project

cargo add axllm

use axllm::{ai, ax};
use serde_json::json;

let llm = ai("openai", json!({"apiKey": std::env::var("OPENAI_API_KEY")?}))?;
let classify = ax("review:string -> sentiment:class \"positive, negative, neutral\"")?;

Set OPENAI_APIKEY in your environment before running provider-backed code.

In the ax repo

From a clone of the ax repo:

npm run example -- rust src/examples/rust/mcp/native_mcp_tools.rs

Active practice

Show that you can use it

Answer 2 in a row to learn this · attempt 1

Keep exploring

See the idea in context

Show that you can use it

Source-backed follow-up