agent-anything

agent-anything is a command-line tool for running AI coding agents and writing their final responses to the terminal or a specified file.

Currently supported providers:

• Codex (default)

Installation

Requirements

• uv is installed
• Codex CLI is installed and authenticated, and the codex command is available in your terminal

Run the following command to verify that Codex CLI is available:

codex --version

Install

uv tool install agent-anything

After installation, run:

agent-anything --help

You can also run it temporarily with uvx without installing it:

uvx agent-anything "Summarize the directory structure of the current project"

Usage

agent-anything "Review the current project and suggest improvements"

Write the result to a file

agent-anything "Generate a code review report for the current project" \
  --output review.md

Output JSON

JSON output requires a JSON Schema provided through --json-schema. You can pass either a JSON object directly or the path to a Schema file.

Pass a JSON Schema directly:

agent-anything "Summarize the current project" \
  --format json \
  --json-schema '{"type":"object","properties":{"summary":{"type":"string"}},"required":["summary"],"additionalProperties":false}'

Use a JSON Schema file:

agent-anything "Summarize the current project" \
  --format json \
  --json-schema schema.json \
  --output result.json

Specify the model and reasoning effort

agent-anything "Analyze this problem" \
  --model <model-name> \
  --reasoning_effort high

Available reasoning effort levels are low, medium, high, xhigh, max, and ultra. Actual support depends on the selected model and Codex CLI version.

Command-line options

agent-anything [--format {text,json}]
               [--output OUTPUT]
               [--json-schema JSON_SCHEMA]
               [--provider {codex}]
               [--model MODEL]
               [--reasoning_effort {low,medium,high,xhigh,max,ultra}]
               [--verbose]
               prompt

Features

• Run AI coding agents non-interactively
• Capture and output the agent's final response
• Support plain text and JSON Schema-based structured output
• Write results to the terminal or save them to a file
• Select the model and reasoning effort for each task