claudeless

A Claude CLI simulator for integration testing tools that invoke claude.

Overview

Claudeless provides a controllable test double that responds to the same CLI interface as the real Claude CLI, enabling deterministic integration testing without API costs.

Installation

Cargo

cargo install claudeless

Using in Test Suites

Add claudeless as a dev dependency to get the binary built alongside your tests:

[dev-dependencies]
claudeless = "0.2"

In your tests, use CARGO_BIN_EXE_claudeless to find the binary:

use std::process::Command;

#[test]
fn test_my_claude_integration() {
    let claudeless = env!("CARGO_BIN_EXE_claudeless");

    let output = Command::new(claudeless)
        .args(["--scenario", "tests/fixtures/scenario.toml"])
        .arg("-p")
        .arg("hello world")
        .output()
        .unwrap();

    assert!(output.status.success());
}

Usage

Run with a scenario file:

claudeless --scenario scenario.toml -p "hello"

Run interactively (TUI mode):

claudeless --scenario scenario.toml

Scenario Files

Control responses with TOML scenario files:

# scenario.toml
default_response = "I'm not sure how to help with that."

[[responses]]
pattern = { contains = "refactor" }
response = "I'll help refactor that code."

[[responses]]
pattern = { regex = "fix bug #(\\d+)" }
response = { text = "Fixed!", delay_ms = 100 }

See the Scenario Reference for full documentation.

Features

• Pattern matching: substring, exact, regex, and glob
• Response delays for timing tests
• Multi-turn conversation simulation
• Tool call simulation
• Failure injection (network errors, rate limits, auth failures)
• Capture log for request inspection
• TUI mode for interactive testing

License

MIT - Copyright (c) 2026 Alfred Jean LLC

"Claude" is a trademark of Anthropic, PBC. This project is not affiliated with Anthropic.