Who it is forA developer who starts many Codex tasks in the same repository and keeps re-explaining the architecture and past decisions.
The problemEach new Codex task starts without the project history, so decisions and context must be repeated.
The outcomeThe project has an EchoesVault folder and a portable runtime, and later Codex tasks show a status card and can restore memory.
What you need first
- Codex in the ChatGPT desktop app or Codex CLI
- Python 3.9 or newer available as python3
- Git, when installing from GitHub
Steps
- Run
codex plugin marketplace add psinetron/echoes-vault-codexonce in a terminal. - Run
codex plugin add echoes-vault-codex@echoes-vault. - Run
codex plugin listand confirm the entry shows installed, enabled. - Start a new Codex task in your project and send: Initialize EchoesVault for this project.
- Commit the generated EchoesVault/, .echoes-vault/echoes_vault.py, and adapter files with Git.
Prompt for your agent
Using the EchoesVault plugin from psinetron/echoes-vault-codex, initialize project memory in <project-path> and list the files it created. Expected output: an EchoesVault/ folder and a .echoes-vault/ folder containing echoes_vault.py. Verify that EchoesVault/.gitignore ignores index.md.
Caveat. Initialization is project-local, and the plugin stays silent in repositories that have not been initialized, so each project needs its own setup.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for psinetron/echoes-vault-codex and "Set up persistent project memory for a repo in Codex"
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer