Another insane Jev use case! Jev is making it dramatically cheaper to evaluate what actually happened inside an agent run. And finally, someone open-sourced a self-improving memory layer that can put that signal to work across agent harnesses: - Claude Code - Codex - Cursor - OpenCode, and 20+ more Beacon by @asymptotelabs continuously captures your agent history across harnesses and uses Jev to identify which runs are actually worth learning from. It then turns the highest-signal workflows, corrections, and debugging patterns into reusable skills. GitHub repo: http://github.com/Asymptote-Labs/agent-beacon… (don’t forget to star it ) Beacon preserves the complete session history. But preserving a run and learning from it are two different things. Most coding-agent sessions contain routine exploration, failed commands, and fixes that only apply to one task. The trace can remain available for inspection without turning every detail into guidance for future agents. Jev scores each run for evidence, reuse potential, and human correction signals. An application policy then decides whether to promote, review, or discard it. The recording shows this in action. Claude receives a coding task, modifies the implementation, and runs the tests. I then provide an edge-case correction, so Claude updates the code and adds regression coverage. Beacon automatically captures the complete session. Jev evaluates whether the correction contains a reusable engineering lesson. Once approved, that lesson becomes available to other coding agents working on the project. Since it works across harnesses: - Claude Code sessions can teach Codex. - Cursor debugging can improve OpenCode. So a problem solved by one agent should not need to be learned from scratch by another. If you want to dive deeper into Jev, I also wrote a hands-on guide to building this Jev-style decision path with open models, entirely locally. Read it below.

Asymptote-Labs/agent-beacon

Source: https://github.com/Asymptote-Labs/agent-beacon

Website · Docs · Discord · Install · Commands

Beacon captures agent session history across Claude Code, Cursor, Codex, OpenCode, and 20+ other harnesses, then turns useful workflows, corrections, and debugging patterns into reusable knowledge for future agents.

A problem solved by one agent shouldn’t need to be learned from scratch by another.

Beacon Overview

Beacon is an open-source memory layer for AI coding agents that learns from your work across Claude Code, Cursor, Codex, OpenCode, and 20+ other harnesses. It captures full session history, identifies useful workflows, corrections, and debugging patterns, and makes that knowledge reusable by future agents. Built for developers who want agent knowledge to compound across tools instead of disappearing when a session ends.

Key Capabilities:

• Cross-harness history - sessions from Claude Code, Cursor, Codex, OpenCode, Cline, and 20+ harnesses in one place
• Knowledge that compounds - workflows, corrections, debugging patterns, and repo conventions that survive beyond a single session
• Shared agent memory - reviewed knowledge future agents can retrieve through MCP or Agent Skills
• Exact session replay - prompts, responses, tools, commands, edits, approvals, MCP activity, and tokens in one trace
• Local-first portability - durable JSONL, explicit destinations, and no harness lock-in

🚀 Quick Start

Beacon is open source and local-first. Interactive endpoint setup signs in through beacon.sh and preselects Beacon Managed, with an explicit Local opt-out. Signing in forwards nothing; confirming Managed installs Beacon and connects this machine in the same command, and the confirm screen says so before you accept. System, package, MDM, and CI installation paths remain noninteractive and account-free.

1. Install Beacon

brew trust asymptote-labs/tap
brew tap asymptote-labs/tap
brew install beacon

beacon endpoint install

sudo apt install ./beacon_<version>_linux_amd64.deb

sudo dnf install ./beacon_<version>_linux_amd64.rpm

msiexec /i BeaconEndpointAgent-<version>-x64.msi

msiexec /i BeaconEndpointAgent-<version>-x64.msi /qn

2. Use your agents normally

Open Claude Code, Cursor, Codex, or any other supported harness.

Beacon continuously captures your session history in the background

3. Explore your history

beacon traces

This opens a local terminal browser for traces, event timelines, token usage, and retained content. Nothing is sent anywhere. To use the local web view instead:

beacon endpoint dashboard

Or inspect the raw event stream:

~/.beacon/endpoint/logs/runtime.jsonl

Signing in does not enable forwarding. Confirming the preselected Beacon Managed option does: the wizard says so on the confirm screen, names what your chosen privacy mode sends, and connects the endpoint after the install succeeds. Choose Local to keep everything on this machine, and disconnect any time with beacon endpoint disconnect.

Inspect the account used during interactive setup:

beacon whoami

🧠 Turn Session History Into Memory

Every agent session contains potentially useful knowledge about your codebase.

Beacon creates a loop around that history:

Run agents
    ↓
Capture session history
    ↓
Evaluate what worked
    ↓
Extract useful knowledge
    ↓
Review + approve
    ↓
Reuse across future agents

That could be:

• the right way to run a migration
• a debugging path that finally fixed an obscure issue
• a testing convention agents repeatedly get wrong
• a repository-specific workflow
• the right sequence of internal tools
• a correction you’ve given multiple agents

Instead of disappearing into old sessions, that knowledge becomes reusable.

🔀 Cross-Harness by Design

Most agent memory belongs to a single harness.

Beacon sits across the harness layer.

Claude Code ─┐
Cursor ──────┤
Codex ───────┼──→ Beacon ──→ shared project knowledge
OpenCode ────┤
Cline ───────┘

Because Beacon captures and normalizes session history across tools, knowledge learned through Claude Code doesn’t have to stay in Claude Code.

Your Cursor sessions can improve Codex.
Your Codex sessions can improve OpenCode.
Your history keeps compounding even as you switch tools.

Your agent session history belongs to you, not the harness.

🔎 One Trace Format for Every Agent

Beacon captures agent execution where it happens and normalizes it into a common OpenTelemetry-based event model.

That includes:

• sessions
• prompts and responses
• tool calls
• commands
• file activity
• approvals
• MCP interactions
• token usage

Instead of separate proprietary histories for every coding tool, you get one dataset you can inspect, search, learn from, and build on.

🖥️ Local Dashboard

Beacon ships with a local, read-only dashboard:

beacon endpoint dashboard

Use it to explore session history across harnesses and understand what your agents actually did.

The underlying JSONL remains directly accessible, so you’re never dependent on the UI.

Supported Agents

Beacon supports local agents, browser agents, cloud agents, CI workflows, and agent SDKs.

Local Agent Coverage

Browser Chat

Cloud Agents

SDK Instrumentation

Output Destinations

Beacon writes endpoint telemetry to local JSONL by default.

You can also forward the same normalized session history into infrastructure you already use:

Splunk · Datadog · Elastic · Microsoft Sentinel · CrowdStrike Falcon LogScale · Sumo Logic · Wazuh · AWS S3 · GCS · CloudWatch

Architecture

Beacon captures activity where agents actually run and normalizes it into one shared event model.

Local agents ───────┐
Browser chat ───────┤
CI ─────────────────┼──→ Beacon ──→ unified session history
Cloud agents ───────┤                    │
Agent SDKs ─────────┘                    ├──→ local JSONL
                                         ├──→ reviewed memory
                                         ├──→ MCP / Agent Skills
                                         └──→ your own infrastructure

See the documentation for the full architecture breakdown.

Documentation

Read the docs for:

• installation
• supported runtimes
• event schema
• session history
• memory
• MCP
• Agent Skills
• forwarding
• advanced configuration

Read the docs →

Contributing

Contributions are welcome.

Open an issue, submit a pull request, or join the Discord.

License

MIT

Star History