LinuxCommandLibrary
GitHubF-DroidGoogle Play Store

ra-aid

AI-powered autonomous development companion

TLDR

Run with a task description
$ ra-aid -m "[add unit tests for auth module]"
copy
Run in research mode to gather information first
$ ra-aid -m "[understand the caching strategy]" --research-only
copy
Use a specific provider and model
$ ra-aid -m "[fix the bug]" --provider [openai] --model [gpt-4]
copy
Run in chat mode for interactive collaboration
$ ra-aid -m "[help me design the API]" --chat
copy
Run with cowboy mode to skip shell command approval prompts
$ ra-aid -m "[fix the linting errors]" --cowboy-mode
copy

SYNOPSIS

ra-aid [options] -m message

DESCRIPTION

ra-aid is an autonomous AI development companion built on LangGraph that operates in three distinct stages: research, planning, and implementation. It analyzes codebases, formulates strategies, and executes multi-step development tasks with minimal human intervention.The tool uses a structured workflow where it first researches the codebase and task requirements, then creates a detailed plan, and finally implements changes. This staged approach provides transparency and allows developers to review plans before execution.ra-aid supports multiple AI providers (Anthropic, OpenAI, OpenRouter, Gemini, and others) and integrates with shell commands, file editing, and web browsing capabilities to accomplish complex development tasks. The --chat mode enables interactive collaboration where you can guide the agent through tasks conversationally.

PARAMETERS

-m, --message MESSAGE

Task description for the agent.
--provider PROVIDER
AI provider to use (anthropic, openai, openrouter, openai-compatible, makehub, gemini).
--model MODEL
Specific model to use (required for non-Anthropic providers).
--research-only
Only gather information without making changes.
--chat
Enable interactive chat mode with direct human interaction (implies --hil).
--hil
Enable human-in-the-loop mode for interactive approval.
--cowboy-mode
Skip interactive approval for shell commands.
--auto-test
Automatically run tests after each code change.
--test-cmd COMMAND
Custom command to run tests.
--recursion-limit N
Maximum recursion depth for agent operations (default: 100).
--show-cost
Display cost information as the agent works.
--max-cost USD
Set a maximum cost threshold in USD.
--verbose
Enable verbose output.
--help
Display help information.
--version
Display version information.

CAVEATS

Requires API keys for AI providers. Complex autonomous tasks may consume significant API tokens. Use --show-cost to monitor spending and --max-cost to set limits. Always review changes made in autonomous mode before committing.

HISTORY

ra-aid was created as an open-source autonomous coding agent built on LangGraph's agent-based task execution framework. It differentiates itself through its three-stage research-plan-implement workflow, offering more transparency than single-pass coding agents.

SEE ALSO

claude(1), aider(1), codex(1), forge(1)

Copied to clipboard
Kai