Praxis: Proof, Memory, and Context for AI Agents

Overview

Praxis is an open-source tool designed to help AI agents work more effectively by keeping track of what they do and why. It is specifically built to work with Claude Code sessions. The main problem it solves is that AI interactions often lack a clear record, and project details get lost when a session ends. Praxis fixes this by creating a system that captures proof of actions and remembers the reasoning behind decisions. It runs entirely on your local machine, meaning your code and private data never leave your computer.

Benefits

Praxis offers several key advantages for developers using AI tools.

First, it provides proof of action. AI agents often say they succeeded in a task, but it can be hard to verify. Praxis creates signed receipts that seal exactly what happened during a session. These receipts are tamper-evident, meaning if someone tries to change the record, the chain breaks and it becomes obvious.

Second, it preserves memory of reasoning. When an AI session ends, Praxis writes down important decisions, constraints, and why certain paths were avoided. This prevents future sessions from wasting time on dead ends or re-inventing solutions that already failed.

Third, it is completely local and private. The system runs on your machine without needing a server or an account. No sensitive data leaves your environment. It also adds zero cost to your existing subscription because it does not make its own API calls.

Finally, it reduces warm-up time. New sessions start with full context because the memory file is automatically read. This means developers do not have to re-explain the project stack or specific rules to the AI every time they start a new task.

Use Cases

Praxis is useful for any developer or team that uses AI agents to write code or manage projects.

It works well for teams that want to share project history. While the memory file is ignored by default in version control, teams can choose to commit it to share context and history across members.

It is ideal for projects where accountability matters. The signed receipts allow anyone to verify offline exactly what an AI agent attempted or did during a session without needing an internet connection or a vendor account.

It helps in complex software development where context is critical. By storing architectural decisions and load-bearing code warnings in a plain Markdown file, it ensures that the AI always knows the rules of the project.

It integrates easily with Git. It records the prompts asked and the reasons behind AI-written commits in Git notes. This turns reviewing AI-generated pull requests from a confusing task into a clear testimony of intent.

Pricing

Praxis is open-source and free to use. It rides on top of your existing Claude Code subscription and does not add any extra API costs or fees.

Vibes

As an open-source project, Praxis has not yet gathered a large volume of public reviews or testimonials. However, the design philosophy focuses on transparency and user control, which has been well-received in developer communities that value privacy and local-first tools.

Additional Information

Praxis is currently available as an open-source project. It requires Node 22 or higher to install. The setup is simple and involves running a single command in the project directory. The Axolotl Companion, a status bar application that tracks session health, is currently available on Windows, with support for macOS and Linux coming soon. The project follows a local-by-design approach, ensuring that all data remains on the user's machine unless they opt in to anonymous usage counts.