@gengdaJ: Genius! Use external Obsidian to permanently preserve Codex's important memories, so Codex won't forget key tasks! I have organized the content you need to send to your Codex into a super prompt: (Remember to specify the Vault path) Please help me configure Obsidian as Codex's cross-project…
Summary
Introduces how to configure Obsidian as Codex's cross-project long-term memory repository, allowing Codex to persistently save important information and avoid forgetting.
View Cached Full Text
Cached at: 05/23/26, 06:15 PM
Brilliant! Use external Obsidian to persistently store Codex’s important memories, so Codex never forgets key things!
I’ve organized what you need to send to Codex into a super prompt (remember to specify the Vault path):
Please help me configure Obsidian as Codex’s cross-project long-term memory.
My Obsidian Vault path is:
<your Obsidian Vault path>
Inside this vault, create a folder: Codex记忆
Target structure:
Codex记忆/
├── AGENTS.md
├── TODO.md
├── agent/
│ └── open-loops.md
├── 项目/
├── 人物/
├── 工作流/
├── 决策/
└── 素材/
Please complete the following:
- Create the above directories and base Markdown files.
- In
Codex记忆/AGENTS.md, write the Codex memory usage rules. - In Codex’s global
AGENTS.mdor custom instructions, append an “Obsidian Codex Memory” rule so that Codex will know this directory is its long-term memory. - Do not modify any content in the global
AGENTS.mdthat is unrelated to Obsidian Codex memory. - Do not save any cookies, tokens, API keys, passwords, verification codes, ID numbers, bank cards, private contact information, or other sensitive data.
- Do not write entire chat logs into the memory; only save information that is long-term useful and will be referenced repeatedly.
- After completion, verify that the files were created successfully, and tell me which files were created and which global instruction file was modified.
Write the following rules into Codex记忆/AGENTS.md:
Codex Memory Usage Rules
This folder is Codex’s cross-project long-term memory. It sits inside an Obsidian vault, so all content should be readable, editable, searchable, and diff‑able Markdown.
When to Read
Before starting a significant or long‑running task, quickly scan this file and check the relevant subdirectory based on the task type:
- Project-related:
项目/ - People, collaborators, user personas:
人物/ - Reusable workflows, commands, checklists:
工作流/ - Important decisions, trade‑offs, rationale:
决策/ - Reusable assets, templates, phrasing:
素材/ - Cross‑project open items:
TODO.mdandagent/open-loops.md
When to Write
Write only when you learn something long‑term useful that will be referenced repeatedly. Prioritise saving:
- User’s long‑term preferences, explicit boundaries, repeatedly emphasised work styles
- Stable project paths, key commands, environment differences, release processes
- Verified root causes, fixes, debugging sequences
- Important decisions and the reasoning behind them
- Cross‑project to‑dos and unresolved questions
- Reusable article structures, script workflows, research templates, prompts
What NOT to Write
- Do not save full chat logs, day‑to‑day logs, temporary emotions, or one‑off intermediate steps
- Do not save cookies, tokens, API keys, passwords, verification codes, ID numbers, bank cards, private contact information
- Do not copy third‑party platform configurations, account credentials, or sensitive values from logs into this folder
- Do not write low‑value summaries just to “look like you have memory”
How to Write
- Prefer updating existing notes; create new notes only when no suitable one exists
- Write short, reviewable Markdown snippets each time
- Use clear headings, dates, source task, and scope
- Separate facts from inferences — avoid solidifying guesses as rules
- If old memory becomes outdated, do not delete it directly; first mark it as “outdated” and explain why
Close‑out Rule
Before ending an important task, perform a memory close-out:
- Determine whether any content with long‑term value needs to be recorded
- If so, update the corresponding file
- Unresolved items go into
TODO.mdoragent/open-loops.md - In the final reply, briefly state which memory files were changed
Append the following content to Codex’s global AGENTS.md or custom instructions:
Obsidian Codex Memory
Use this directory as a cross‑project long‑term memory:
<your Obsidian Vault path>/Codex记忆
Before starting a significant or long‑running task, quickly browse:
<your Obsidian Vault path>/Codex记忆/AGENTS.md
When you learn something long‑term useful that will be referenced repeatedly, update the relevant Markdown files inside <your Obsidian Vault path>/Codex记忆. Focus on saving: stable project paths, key commands, user’s long‑term preferences, explicit boundaries, verified debugging conclusions, important decisions, reusable workflows, and cross‑project open items.
Do not save full chat logs, day‑to‑day logs, one‑off intermediate steps, or treat temporary processes as memories. Do not save cookies, tokens, API keys, passwords, verification codes, ID numbers, bank cards, private contact information, or copy third‑party platform configurations or sensitive values from logs into Obsidian memory.
Writing rules: Prefer updating existing notes; create new notes only when no suitable one exists. Write short, reviewable Markdown snippets each time. Separate facts from inferences. If old memory becomes outdated, do not delete it directly; first mark it as “outdated” and explain why.
Before ending an important task, perform a memory close-out: determine whether any content with long‑term value needs to be recorded; if so, update the corresponding file; unresolved items go into <your Obsidian Vault path>/Codex记忆/TODO.md or <your Obsidian Vault path>/Codex记忆/agent/open-loops.md; in the final reply, briefly state which memory files were changed.
Dan McAteer (@daniel_mac8):
How to setup persistent Codex memory in Obsidian
Copy the below prompt into Codex. The prompt instructs Codex to create the memory folders in Obsidian
Copy the below custom instructions. They instruct Codex to use Obsidian Codex memory to save memories
That’s it!
Similar Articles
@Saccc_c: Codex Advanced Memory Techniques: Building Cross-Project Shared Memory with Obsidian Knowledge Base Basic configuration is two steps: 1. Copy the prompt from Figure 1 to Codex, letting it create a memory folder in Obsidian 2. Copy the instructions from Figure 2 and add them to the custom instructions in settings, as part of the global AGENTS.md…
Introduces how to use the Obsidian knowledge base to build cross-project shared memory for Codex, automatically depositing valuable context by copying prompts and global instructions.
@daniel_mac8: How to setup persistent Codex memory in Obsidian 1. Copy the below prompt into Codex. The prompt instructs Codex to cre…
A guide on setting up persistent memory for Codex using Obsidian by copying a prompt and custom instructions for the AI assistant.
@Saccc_c: Codex's Memories feature is a setting you must enable; it gives Codex active memory. How to enable: Settings → Personalization → Memory → Enable Memory (Figure 1). After enabling and letting it run for some time, you can open memory.m...
Introduces Codex's Memories feature, which gives Codex active memory, along with instructions on how to enable it and how to view the memorized content.
@op7418: Codex keeps getting better—new Chronicle memory continuously takes local screenshots so it knows which doc or bug you mean without you spelling it out. Pro users only for now.
Codex adds Chronicle, a local screenshot-based memory feature that lets it track open docs and bugs without explicit context.
@yanhua1010: For most people, Obsidian eventually becomes a beautifully organized filing cabinet. It stores a lot, but very little is actually used. This is because it is designed primarily for "input": clipping, organizing, tagging, and folders. However, a truly useful knowledge base should proactively push insights back to you every day. I created an auto-growing Obsidian knowledge base Demo: …
The author shares an automated Obsidian knowledge base Demo that, by integrating with Claude, delivers daily briefings and facilitates knowledge compounding, transforming passive storage into active insight.