Star Beacon on GitHub →

Star Beacon is an open-source memory layer for AI coding agents that learns from user work across various tools. It captures full session history, identifies useful patterns, and makes that knowledge reusable for future agents. It works with Claude Code, Cursor, Codex, OpenCode, and over 20 other harnesses.

Beacon is installed locally and does not send any data anywhere. It opens a local terminal browser for session history, token usage, and retained content. It runs in the background and never forwards data unless a user signs in to the Beacon Cloud option. Signing in installs Beacon and starts forwarding session history to the cloud.

The system installs the latest Beacon package using the package manager for the user's distribution and architecture. Users can specify a particular release version if needed. Beacon also installs Beacon Skills in specified directories unless turned off. Users can manage these skills later.

Beacon captures agent execution and normalizes it into a common OpenTelemetry-based event model. This creates a single dataset that can be inspected, searched, learned from, and built upon, regardless of the tool used. Beacon provides a local, read-only dashboard for exploring session history across different harnesses.

Beacon supports various types of agents, including local, browser, cloud, CI workflows, and agent SDKs. It provides commands for installing skills in different runtimes. By default, Beacon writes endpoint telemetry to local JSONL files. Users can also forward this data to the cloud.