Summary

OpenChronicle offers a local-first, inspectable memory layer for tool-capable LLM agents, serving as an open alternative to proprietary systems. It enables AI agents to leverage real-world working context, providing a hackable, model-agnostic solution that enhances agent utility and transparency for developers.

How It Works

The system captures structured context primarily from macOS Accessibility Tree (AX Tree) events, supplemented by screenshots. This AX-first approach prioritizes lower processing costs, better intent capture, and a smaller memory footprint. Captured data is processed through event dispatching, parsing, normalization into sessions, and reduction before local storage as human-readable Markdown and indexed SQLite FTS5 for agent querying.

Quick Start & Requirements

• Install: Clone (git clone https://github.com/openchronicle/openchronicle.git), cd, then bash install.sh.
• Prerequisites: macOS 13+ and Xcode Command Line Tools (xcode-select --install).
• Run: openchronicle start launches the memory daemon.
• Agent Connection: MCP clients connect to http://127.0.0.1:8742/mcp. See docs/mcp.md for details.

Highlighted Details

• AX Tree-first capture for accurate, compact, low-cost memory.
• Local storage combines Markdown with SQLite FTS5 indexing.
• Model-agnostic support via LiteLLM compatibility (Ollama, LM Studio, OpenAI, Anthropic, etc.).
• Event-driven capture from macOS AX events.
• Session-aware memory writing reduces noise.

Maintenance & Community

Currently v0.1.0 and in early alpha. Acknowledges contributors via an "all-contributors" specification. Specific community channels or roadmap details are not provided in the README snippet.

Licensing & Compatibility

Released under the permissive MIT license, allowing broad compatibility, including commercial use and integration within closed-source applications.

Limitations & Caveats

The project is macOS-only and in early alpha (v0.1.0), indicating potential instability. Reliance on AX Tree context may limit effectiveness in highly visual or non-standard application interfaces.