equality-machine/claude-p

claude-p is a command line tool and Python library that lets people use automated, script-friendly access to Claude Code, Anthropic's coding assistant, through the same login session as the interactive app they already use. It is built to solve a specific gap: in some setups, the interactive Claude Code app works fine with a subscription login, but the separate programmatic mode normally used for scripts and automation is limited, capped, or unavailable. Instead of calling that programmatic mode directly, claude-p starts the regular interactive Claude Code application in the background inside a virtual terminal, gives it a specific session id, waits for it to finish responding, and then reads the response from Claude Code's own session log file rather than trying to capture what appeared on screen. The project's README explains that reading this log file is more reliable than scraping terminal output, since terminal displays can drop or garble characters due to cursor movement and redrawing. The tool is designed to be a near drop-in replacement for the existing programmatic command, matching many of the same command line flags and output formats, including plain text, JSON, and a streaming JSON format. It also ships a Python library shaped similarly to Anthropic's official Agent SDK, so a developer can write code that sends a prompt and reads back a response using familiar patterns. Installing it is done through either the uv tool or a standard pip install, and it requires Python 3.10 or newer. The README is direct about its limits. It states clearly that this is a developer convenience tool, not a replacement for the official Anthropic API, and that it does not bypass authentication, billing, or rate limits since it still relies on the user's existing authenticated session and account restrictions. Some features, like replaying certain event types or accepting streaming input, are not yet fully implemented. It is released under the MIT license, a permissive open source license, and is distributed through PyPI.