---
type: "post"
id: "avenoxbeyin-gives-claude-code-codex-and-antigravity-persistent-5a0e74"
slug: "avenoxbeyin-gives-claude-code-codex-and-antigravity-persistent-5a0e74"
title: "avenoxbeyin gives Claude Code, Codex, and Antigravity persistent cross-session memory via an Obsidian vault"
description: "AvenoxBeyin is an open-source system that stores selected project knowledge in local Markdown files and makes it available across supported coding-agent sessions."
retrieval_nugget: "AvenoxBeyin is an open-source system that stores selected project knowledge in local Markdown files and makes it available across supported coding-agent sessions. Its practical distinction is a shared, inspectable vault rather than automatic retention of entire conversations, giving teams a way to carry decisions between clients while keeping the source notes visible."
published_at: "2026-11-14"
updated_at: "2026-11-14"
record_date: "2026-11-14"
date_kind: "published_at"
topics: ["codex","claude","claude-code"]
entities: ["AvenoxBeyin","Avenox","Claude Code","Codex","Google Antigravity","Obsidian"]
related_patterns: ["company-memory-needs-write-loops"]
novelty_label: "new_fact"
what_changed: "AvenoxBeyin adds an Obsidian-vault approach to persistent cross-session memory, supporting Claude Code, Codex, and Antigravity."
prior_refs: ["https://newruntime.com/posts/persistent-memory-for-claude-code-with-weaviate-engram-75aa8e/"]
source_url: "https://github.com/avenoxai/avenoxbeyin"
source_urls: ["https://github.com/avenoxai/avenoxbeyin"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:avenoxbeyin-gives-claude-code-codex-and-antigravity-persistent-5a0e74"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/avenoxbeyin-gives-claude-code-codex-and-antigravity-persistent-5a0e74/","markdown":"https://newruntime.com/posts/avenoxbeyin-gives-claude-code-codex-and-antigravity-persistent-5a0e74.md","json":"https://newruntime.com/posts/avenoxbeyin-gives-claude-code-codex-and-antigravity-persistent-5a0e74.json"}
markdown_url: "https://newruntime.com/posts/avenoxbeyin-gives-claude-code-codex-and-antigravity-persistent-5a0e74.md"
related_shifts: ["shared-context-infrastructure"]
---

# avenoxbeyin gives Claude Code, Codex, and Antigravity persistent cross-session memory via an Obsidian vault

## Retrieval answer

AvenoxBeyin is an open-source system that stores selected project knowledge in local Markdown files and makes it available across supported coding-agent sessions. Its practical distinction is a shared, inspectable vault rather than automatic retention of entire conversations, giving teams a way to carry decisions between clients while keeping the source notes visible.

## Summary
AvenoxBeyin is an open-source system that stores selected project knowledge in local Markdown files and makes it available across supported coding-agent sessions. Its practical distinction is a shared, inspectable vault rather than automatic retention of entire conversations, giving teams a way to carry decisions between clients while keeping the source notes visible.

## What happened
The AvenoxBeyin GitHub repository describes the project as a local memory system for AI clients, including Claude Code, Codex, Google Antigravity, Kiro, Hermes Agent, OpenCode, OMP/Oh My Pi and Pi. It requires Python 3.11 or later and an already installed, configured AI client. The project is MIT-licensed. Obsidian is optional: users can open the vault in Obsidian or another text editor.

The memory consists of Markdown notes in a folder chosen by the user. The project says its index is local and that it supplies selected context to the AI client. It does not automatically turn the full chat history into persistent knowledge; users need to ask the agent to record information they want to keep. A later session can search for a decision or project detail and open the note that contains it. This makes the underlying file, rather than a hidden transcript store, the inspectable record.

The setup can be run through an AI client by asking it to read the project’s setup instructions and install the system in the chosen folder. The repository also documents a manual route using a versioned ZIP and matching SHA-256 file, followed by an installer script. It advises users to review the client’s folder or hook trust prompts and, after installation, test a write and retrieval in a new session. Client integrations differ: for example, the guide tells Codex users to inspect new hook definitions, and it cautions that Codex Desktop skill use is not the same as automatic hook delivery.

The system includes four skills: beyin for finding and saving information and creating tasks or work results; beyin-doktor for checking setup and connections; beyin-guncelle for updates and rollback; and beyin-kontrol for checking whether stored information remains accurate. The last skill is described as running only when requested and not writing without approval. Optional utilities include a capture inbox and a local web view. The repository says the web view is read-only, runs on the same computer, and closes after 30 minutes of inactivity. Basic local checks do not call a model; agent work and the context supplied to a client remain subject to that client’s normal usage and provider terms.

## What the archive adds
New Runtime’s earlier coverage of [Weaviate Engram](https://newruntime.com/posts/persistent-memory-for-claude-code-with-weaviate-engram-75aa8e/) described project history as cross-session context for Claude Code. AvenoxBeyin is a new event in the same memory-for-coding-agents thread: it emphasizes a local Markdown vault shared across several clients, with explicit note-writing and source-file retrieval.

## Trend signal
This supports the assumption that developers will keep using several AI clients on the same project rather than standardizing on one vendor’s integrated memory. A shared vault is one implementation of that direction, though the project’s documentation does not establish how well the approach works across teams. See [Shared context infrastructure](https://newruntime.com/shifts/shared-context-infrastructure/).

## Why now
The repository presents cross-client continuity as a use case: notes, decisions and tasks remain in one folder when a user changes AI clients. That is directly relevant as the documented integrations span multiple coding agents. The source does not explain a particular launch date or external trigger for the release.

## Why it matters
Persistent memory is useful only if builders can tell what was remembered and correct it when it becomes stale. A Markdown vault makes the record inspectable and portable, while the explicit instruction to save important information avoids implying that a whole conversation is reliably absorbed. Teams can use the same project notes with different clients without treating any one vendor’s conversation history as the system of record.

That design also creates operational responsibilities. A mistaken or outdated note can shape later agent work, so teams need review habits and a way to correct or remove bad information. The project’s manual verification step—write a test fact, then retrieve it in a new session—offers a small deployment check. It does not replace checking whether the client’s hooks, folder access and provider data handling match the team’s policies.

## Limits and uncertainty
The repository describes the architecture and integrations, but it does not establish independent measurements of retrieval accuracy, note quality, or time saved. The documentation explicitly says local memory does not mean all AI processing is offline: selected context goes to the AI client, whose provider and account terms apply. Search is based on local keyword matching, and the project does not guarantee that every question will retrieve the right note. Client hook and trust flows differ, and some automatic delivery paths are not verified. The project’s stated safeguards and behavior have not been independently audited here.

## Skill to build now
Choose one project decision that should survive a client switch. Ask an agent to save it as a dated or otherwise clearly attributable Markdown note, then start a fresh session and ask for the decision and its source file. Check that the retrieved answer matches the note. This tests both the write path and the cross-session lookup before relying on the vault for higher-stakes project context.

## Reader takeaway
AvenoxBeyin offers a portable memory layer for several coding agents, but it is a notes workflow, not automatic recall of every chat. Its strongest practical test is simple: save a decision, switch to a fresh session, and verify that the agent can retrieve the correct source note.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Codex](https://newruntime.com/topics/codex.md): Explore the codex topic hub.
- topic hub: [Claude](https://newruntime.com/topics/claude.md): Explore the claude topic hub.
- topic hub: [Claude Code](https://newruntime.com/topics/claude-code.md): Explore the claude-code topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)