尚未检测到更新。GitHub 发布版本后会显示在这里。
个人 GitHub 账号下公开仓库的所有者可以登录并验证所有权,自行管理介绍,无需管理员批准。
登录并生成验证文件,在 30 分钟内将其提交到此仓库的默认分支。其他分支或分叉仓库中的文件无效。然后在下方检查文件,即可无需管理员批准直接管理项目。请保留文件以维持管理权限。
如需登录,将前往登录页面。登录后请回到此处,再次点击所选的验证按钮。
在项目文件夹中运行 /nomorevibe verify,即可验证所有权并自行管理介绍。
一分钟完成验证Long-term memory for OpenAI Codex (CLI, IDE, Desktop) that follows you to Claude Code, Cursor and your other machines. Automatic recall and capture; works alongside AGENTS.md and Codex's own local memories. By Atlaso.
翻译准备中
取自仓库 README
正在加载完整 README。
查看完整 README ↗Automatic long-term memory for OpenAI Codex. Atlaso recalls relevant context before each prompt, captures durable facts and decisions as you work, and syncs across your sessions and devices — for the CLI, the IDE extension, and the Desktop app (they share one config, so a single install covers all three).
Codex reads AGENTS.md and keeps its own local memories on the machine you run it on. This connector gives Codex CLI, IDE and Desktop a memory that carries yesterday's work into today, your shell and tooling choices, and the decisions you don't want re-litigated, and shares it with Claude Code, Cursor and your other tools. Automatic in Codex CLI; Codex Desktop installs the same plugin from its marketplace. How it works · Docs
Install uv first. The plugin runs on a uv-managed Python, so uv is the only prerequisite. Without it, prompts run without recall and Codex shows "Atlaso needs uv. Install uv, then retry your prompt; memory will set up in the background."
atlaso-labs/codexmainatlaso setup prepares the runtime up front and atlaso status says whether your next prompt gets recall.codex plugin marketplace add atlaso-labs/codex
codex plugin add atlaso@atlaso
| Piece | Mechanism | What it does |
|---|---|---|
| recall | UserPromptSubmit hook | injects recalled memory before Codex sees your prompt |
| capture | Stop hook | saves the finished exchange (instant local write, synced later) |
| start | SessionStart hook | eligible personal + current-project Ambient Memory, then background cache sync |
| memory tools | MCP server — recall, remember, forget, recent, status | deliberate, model-invoked memory |
| skill + rules | SKILL.md + AGENTS.md | when to remember, personal vs project, fixing memory |
Recalled memory is injected as a plain branded block — no instructions, the model decides how to use it:
=== Atlaso Memory ===
- recalled note
- recalled note
=== Atlaso Memory ===
Hooks fail open: any error → silent no-op, your turn proceeds. Free plan = local-only; paid = cloud sync across devices.
Memory is stored locally on your machine by default. Cloud sync is opt-in (sign in). Learn more at atlaso.ai.
A past Desktop update temporarily stopped hooks from firing (openai/codex#21639). If automatic recall/capture seems quiet on Desktop, the model-invoked memory tools (MCP) still work. Set ATLASO_DEBUG=1 and check ~/.atlaso/atlaso-codex-*.log to confirm hooks fire.
Built by Atlaso Labs Inc. · source: github.com/atlaso-labs/codex
When enabled on an eligible plan, the start hook requests a brief for this tool and the host event's project directory. Personal context can accompany the current project's context. Missing or untrustworthy project identity requests personal context only. A first session can receive a brief without a local cache. Each session validates the current server policy; disabled, offline, rejected, or incompatible responses inject no Ambient Memory. Reopening after recovery retries automatically. Per-turn recall remains a separate feature.
The Python start path has a three-second budget on POSIX; the host hook manifest sets a five-second timeout, including runtime startup. The latter is also the Windows bound. First-time runtime installation can exceed that budget and yield no brief until dependencies are available. These source changes require publishing the rebuilt plugin and updating the installed version; a source checkout alone does not update a user's plugin. Host versions must actually execute SessionStart.