Yanzi

Links: yanzi.io | chucksailer.me Agent Setup: Tell your AI Agent to use Yanzi (Codex, Copilot, etc.)

AI-assisted development generates decisions and reasoning that are often lost across chat sessions, commits, and ad hoc notes. Git captures code changes, but not the full decision trail behind those changes. Yanzi provides deterministic logging for AI-assisted development so decisions can be recovered, audited, and shared.

To get started with your favorite AI Agent, just copy the Agent Boot Strap Prompt and paste into your AI Agent.

What Yanzi Is

Yanzi is a cross-platfor (Windows, Mac OS, and Linux) CLI written in golang and is designed to be used by AI Agents. Yanzi is not an MCP. Yanzi is a CLI and can be queried by a human from the command line. Yanzi is a deterministic logging layer for AI-assisted development.

It records:

• Captures (prompt/response records)
• Checkpoints (milestone summaries)
• Role/meta events (agent control intent)
• Optional metadata for capture context

Yanzi allows you to pause this capture as well. Yanzi offers a variety of commands, known as @yanzi commands, that allow both the AI Agent and Human to interact with Yanzi more easily. Yanzi allows exports of the event stream, in structured formats, so AI-assisted work can be piped into other systems for later review or analysis.

Yanzi is not an orchestration framework or automation engine.

Installation

Install Yanzi with Go:

go install github.com/chuxorg/yanzi/cmd/yanzi@latest

Binary downloads are also available from GitHub Releases.

Quick Start

yanzi project create "alpha"
yanzi checkpoint create --summary "Initial state"

During development, agents can record captures and checkpoints as work progresses, then export logs when needed.

Agent Protocol

Yanzi is designed for AI-agent workflows that use explicit control lines, for example:

@yanzi role Engineer
@yanzi checkpoint "Refactor authentication flow"
@yanzi pause
@yanzi resume
@yanzi export

Agents translate these protocol lines into Yanzi CLI commands.

Exporting Logs

Supported export formats:

yanzi export --format markdown
yanzi export --format json
yanzi export --format html

• Markdown: human-readable log
• JSON: structured machine-readable log
• HTML: professional presentation of the log

Using Yanzi Logs in External Systems

Yanzi JSON export can be ingested by external systems such as ELK, Splunk, Sentry, or other logging pipelines.

Yanzi writes JSON to ./YANZI_LOG.json by default. Example usage patterns:

yanzi export --format json > log.json

yanzi export --format json | curl -X POST https://example.internal/logs --data-binary @-