Developer Tools · Reviewed November 16, 2026
Beacon open source coding agent memory brand logo mark

Beacon

Beacon is open-source memory for coding agents that records sessions across Claude Code, Cursor and Codex and turns them into reusable skills.

Pricing
Freemium
Rating
4.85/ 5 · 106 reviews
Last reviewed
November 16, 2026
Channels
Beacon homepage presenting self-improving memory for coding agents with install commands
01

Overview

Beacon

Beacon is an open-source memory layer for AI coding agents, built by Asymptote Labs. Beacon records your agent sessions across Claude Code, Cursor, Codex, OpenCode, Cline, GitHub Copilot and more than 20 other agent tools in one timeline: prompts, tool calls, file edits, commands, approvals, MCP activity and token usage. A process the company calls dreaming then reviews successful sessions, picks out the correction or workflow that mattered, and drafts it as a skill. Once you approve it, any agent can pull that skill over MCP or as an Agent Skill, so a fix one agent learned does not have to be relearned by the next.

Production credibility: Beacon's code is on GitHub under the MIT licence in the Asymptote-Labs/agent-beacon repository, written mainly in Go, with more than 3,500 stars in mid-November 2026. Asymptote Labs announced a $5.5 million seed round alongside the launch, and lists security and engineering leaders from Anthropic, Vercel, Modal and Socket among its backers. Beacon was the #1 Product of the Day and #1 Product of the Week on Product Hunt in early November 2026. The open-source CLI records locally by default and uploading is opt-in. Pricing on its own page: Beacon Cloud is free with 50 findings; Pro is $20 a month, or $15 a month billed annually, and adds unlimited findings, full MCP access and agent memories; Team is priced per seat for up to 20 seats in a shared workspace; Enterprise is custom, with SSO, RBAC, audit logs and fleet rollout through Jamf, Intune or Fleet.

Key Features

  • One timeline of sessions from Claude Code, Cursor, Codex, OpenCode and 20+ other agent tools
  • Dreaming process that turns successful sessions into reviewable, reusable skills
  • Skills and past sessions served to any agent over MCP or as Agent Skills
  • Session handoff, so work started in one agent can continue in another
  • Findings that flag risky actions, repeated workflows and costly habits
  • MIT-licensed CLI for macOS, Windows and Linux, plus a TypeScript OpenTelemetry SDK

Ideal Use Case

Developers who move between Claude Code, Cursor and Codex and are tired of explaining the same repository conventions and past fixes to each one. It also suits engineering and security teams that want one record of what their coding agents did, with flags for secrets pasted into prompts, destructive commands and unknown MCP servers, and an Enterprise option to block risky actions.

How Beacon differentiates

Mem0 and Zep are memory APIs that developers build into their own AI apps; Beacon works on the coding agents you already use and learns from their recorded sessions instead. Letta builds its own memory-first agents, while Beacon stays neutral across vendors. Agent observability tools such as Laminar trace agents you write in code; Beacon's focus is the off-the-shelf coding agents on a developer's machine, and turning those traces into skills. Its skill drafts still need a human review before agents use them.

FAQ

Q: What is Beacon? A: Beacon is an open-source memory layer for AI coding agents from Asymptote Labs. It records sessions across Claude Code, Cursor, Codex and more than 20 other agent tools and turns useful workflows and fixes into reusable skills.

Q: Is Beacon free and open source? A: Yes. The Beacon CLI is open source under the MIT licence and can run fully locally. Beacon Cloud has a free plan with 50 findings; Pro costs $20 a month, or $15 a month billed annually.

Q: Which coding agents does Beacon support? A: Beacon lists Claude Code, Cursor, Codex, OpenCode, Cline and GitHub Copilot among more than 20 supported agent tools, across local machines, CI jobs and cloud agents.

Q: Does Beacon upload my code sessions? A: Not by default. Beacon records sessions locally and uploading to Beacon Cloud is opt-in. A Local setup keeps everything on your machine with a local dashboard.

Q: How do agents use what Beacon learns? A: Beacon drafts skills from successful sessions. After you review one, agents can retrieve it over MCP or as an Agent Skill when the same problem comes up again.

tl;dr

Beacon is an MIT-licensed memory layer for AI coding agents that records sessions across Claude Code, Cursor, Codex and 20+ other tools and turns them into reusable skills. It runs locally for free; Beacon Cloud Pro is $20 a month.

02

Key Features

01
Open Source
03

Why Use Beacon

Rating
4.85
Across 106 verified reviews
Saved
282
By ToolDirectory readers
Pricing
Freemium
Publisher-listed pricing model
Listed
Since 2026
Continuously re-reviewed by editors
Category
Developer Tools
Primary listing
Verified by editors during the most recent review · ToolDirectory.AI
Beacon homepage presenting self-improving memory for coding agents with install commands
04

User Reviews

4.85
Out of 5 · 106 ratings
5
94
4
9
3
2
2
1
1
0
05

Similar Tools

Sign up for our newsletter

Receive weekly updates so you can stay up-to-date with the world of AI