Beacon

Beacon captures your coding agent sessions across Claude Code, Cursor, Codex and 20+ tools, then turns the best workflows and fixes into reusable skills — so every agent starts smarter than the last.

Development Tools
AI CodingAgent MemoryOpen SourceDeveloper ToolsClaude CodeCursor
Beacon

Beacon Review: Open-Source Memory for AI Coding Agents That Learns From Every Session

Every AI coding agent starts from zero. You open Claude Code, Cursor or Codex, explain your project structure, your debugging preferences and the fixes you already made yesterday — and tomorrow you explain it all over again. Every correction, every dead end, every "no, not like that" vanishes the moment the session ends.

Beacon is an open-source, self-improving memory layer that fixes this. It captures your agent sessions across 20+ coding harnesses and turns the highest-signal workflows, corrections and debugging patterns into reusable skills, so every agent you use starts smarter than the last.

What is Beacon?

Beacon is a free, open-source memory infrastructure for AI coding agents, built by Asymptote Labs, a New York startup founded by former Doppel engineers Justin D'Souza and Shukan Shah. In October 2026 the company announced a $5.5M seed round led by Tower Research Ventures to build Beacon in the open.

Instead of each coding agent keeping its own isolated history, Beacon sits above the harness layer. It records sessions from Claude Code, Cursor, Codex, OpenCode and 20+ other tools, normalizes them, and uses Jev to identify which runs are actually worth learning from — then distills those into skills your agents can reuse.

Key Features

1. Cross-Harness Session Memory

Beacon captures session history across Claude Code, Cursor, Codex, OpenCode and 20+ coding harnesses. Switch tools mid-task without re-explaining your project: a fix you made in Claude Code becomes context Cursor can use the next day.

2. Self-Improving Skills With Jev

Not every session is worth remembering. Beacon uses Jev to evaluate agent runs cheaply at scale and flag the high-signal corrections, debugging patterns and workflows — then turns them into reusable Agent Skills. You review and approve what gets kept.

3. Runs Locally, Open Source

Beacon is free and open source under the MIT license, with the source code public on GitHub. It runs locally on your machine by default, so your session history stays under your control.

4. One-Command Install

Get started in minutes. On macOS, install with Homebrew (brew tap asymptote-labs/tap && brew install beacon), or run the install script with curl. No credit card required.

5. Built for Teams That Live in Agents

Beacon is designed for developers who already work with AI coding agents every day. It targets the real bottleneck of agent-assisted development: not code generation, but the context that gets lost between sessions.

How It Works

Beacon records your agent sessions as they happen, then evaluates them to find the runs that contain reusable knowledge. High-signal corrections and debugging patterns are distilled into skills, which are fed back to your agents through MCP or Agent Skills. The next time an agent faces a similar problem, it starts from what your team already learned instead of from zero.

Who Should Use Beacon?

  • Developers using Claude Code, Cursor or Codex daily: Stop re-teaching your agents the same project context.
  • Teams switching between coding agents: Keep knowledge portable when you move between harnesses.
  • Open-source contributors: A free, MIT-licensed memory layer you can inspect and extend.
  • Engineering leads: Turn individual debugging experience into shared, reusable team knowledge.

Beacon Pricing

Beacon is free and open source. The core memory layer is free to install and use, with no credit card required. The project is MIT licensed and the source code is available on GitHub.

Beacon Pros and Cons

Pros

  • Free and open source under the MIT license.
  • Works across Claude Code, Cursor, Codex and 20+ harnesses.
  • Turns past sessions into reusable skills instead of just logging them.
  • Runs locally by default.
  • Backed by a $5.5M seed round and an active open-source community.

Cons

  • Still early: the project launched in late 2026 and is evolving quickly.
  • Most valuable for developers who already use AI coding agents heavily.

Final Verdict

If you write code with AI agents every day, Beacon solves the most annoying part of the workflow: starting from zero in every session. It is free, open source and local-first, and it was the #1 Product of the Day on Product Hunt on November 2, 2026.

For developers who switch between Claude Code, Cursor and Codex, Beacon is the missing memory layer — worth installing today.

Try Beacon →

ALTERNATIVE PRODUCTS