Vertiso Memory for Codex

Your memory, inside Codex.

Two commands install the plugin for Codex CLI. The ChatGPT desktop app, Codex CLI, and the Codex IDE extension share one MCP configuration.

Get started → Jump to install ↓

Runs where your agents already work

One memory, every agent

Decide it anywhere. Codex already knows.

What you settle in one tool is on hand in Codex, typed and waiting for your confirmation.

Claude Desktop 9:41 AM
You Heads up: the billing cutover moves to Oct 14. Stripe webhooks stay on v2 until then.
Claude Noted. I saved that as a decision with the v2 constraint attached, dated today.
Remembered · decision · constraint
Vertiso Same room
Codex · new session 9:42 AM
You Start the webhook migration branch.
Codex Your memory says the cutover is Oct 14 and v2 stays live until then. I'll scaffold the v3 handler behind a flag and leave v2 untouched.
Recalled · decision · nothing re-explained

0

times you repeat yourself

8 types

decision, constraint, action, and more

You confirm

nothing closes itself

Install in Codex

The same steps as the install guide, for this tool only: a live Marketplace listing where one exists, the public plugin repository, native HTTP MCP, then the local vmem bridge.

Vertiso Memory MCP URL 
https://memory.vertiso.ai/mcp

OpenAI now provides local Codex work through the ChatGPT desktop app, alongside Codex CLI and the Codex IDE extension. Choose the path for the surface you use.

ChatGPT desktop Work/Codex and Codex CLI: complete plugin

Until Vertiso Memory appears in OpenAI's Plugins Directory, install the complete plugin from its public GitHub repository. It adds the remote MCP server plus checkpoint, handoff, handoff-resume, and wrap-up skills.

OpenAI plugin from GitHub (recommended) 
codex plugin marketplace add Vertiso/memory-codex
codex plugin add vertiso-memory@vertiso-memory-codex

The CLI command installs the plugin for Codex CLI. For the ChatGPT desktop app, restart it, choose Work mode or Codex, open Plugins, select the Vertiso marketplace, and install Vertiso Memory if it is not already shown under Installed. Source and release notes: Vertiso/memory-codex. Full plugins are not available in the Codex IDE extension; use the shared MCP setup below instead. When the directory listing becomes public, switch to that installation so OpenAI manages future updates.

ChatGPT web and iOS: MCP tools

Set up the connection once on ChatGPT web. Open Settings → Security and login, enable Developer mode, then open Settings → Plugins, select +, and create a developer-mode MCP app.

ChatGPT web developer-mode app (OAuth, MCP only) 
Name:  Vertiso Memory
URL:   https://memory.vertiso.ai/mcp
Auth:  OAuth

Complete the browser sign-in. This path exposes Vertiso Memory's MCP tools, not the four bundled skills. After setup on web, this MCP app was verified in the current ChatGPT iOS app. If it does not appear there, refresh or reconnect it from ChatGPT web. Developer-mode app availability can depend on your plan and workspace policy.

ChatGPT desktop, Codex CLI, and Codex IDE extension: MCP tools

These local Codex clients share the same MCP configuration. In the ChatGPT desktop app, open Settings → MCP servers → Add server, choose Streamable HTTP, enter the name and URL, save, restart, then select Authenticate.

ChatGPT desktop / Codex MCP server (OAuth, MCP only) 
Name:  Vertiso Memory
URL:   https://memory.vertiso.ai/mcp

The ChatGPT desktop app, Codex CLI, and Codex IDE extension share ~/.codex/config.toml. To configure it from the CLI, add the server and sign in:

config.toml remote server (OAuth, Codex CLI and IDE) 
# ~/.codex/config.toml
[mcp_servers.vertiso-memory]
url = "https://memory.vertiso.ai/mcp"
# then: codex mcp login vertiso-memory

In the Codex IDE extension, open the gear menu → MCP servers → Add server, enter the name and URL, choose Streamable HTTP, save, restart the extension, then select Authenticate.

Need a bearer token or the local vmem bridge?

Without browser OAuth, the Node-based mcp-remote bridge stores the bearer token in the shared Codex configuration:

~/.codex/config.toml (bearer token fallback) 
[mcp_servers.vertiso-memory]
command = "npx"
args = [
  "-y",
  "mcp-remote",
  "https://memory.vertiso.ai/mcp",
  "--header",
  "Authorization: Bearer <paste your token here>",
  "--header",
  "X-Agent-Source: agent:codex",
]

For a local bridge that keeps the token in ~/.vmem, run vmem mcp install codex. Restart any open ChatGPT desktop, Codex CLI, and Codex IDE extension sessions afterward.

~/.codex/config.toml (vmem bridge) 
[mcp_servers.vertiso-memory]
command = "<run `which vmem` to get the absolute path>"
args = ["mcp", "serve", "--client", "codex"]

Recommended: add a bootstrap instruction to ChatGPT custom instructions or your AGENTS.md for local Codex work so your assistant registers with Vertiso Memory at the start of every session:

Bootstrap instruction 
Vertiso Memory (vmem) is my persistent memory across every tool and session.

- Call the `hello` tool at the start of every session. Treat its primer
  as authoritative context, not a suggestion.
- `recall` before planning or answering anything that may depend on
  earlier context. Look it up instead of asking me to repeat it.
- `remember` durable decisions, preferences, constraints, and project
  state the moment they land. `update_memory` rather than writing a
  near-duplicate.
- Offer a handoff before the session ends, compacts, or moves to
  another tool.

Connected? Start a new conversation and say Hello. Your assistant greets you by name and mentions what it remembers. The standing instruction and the first conversation →

API token

Only the vmem bridge and the bearer-token fallbacks need one. OAuth clients sign in during the handshake instead.

Log in to get your token →

Power users: the vmem CLI

A native binary for working memory from the terminal, and a stdio-to-HTTP bridge for clients that cannot reach a remote server. It keeps your token in ~/.vmem instead of each app's config.

Install vmem 
curl -fsSL https://memory.vertiso.ai/install.sh | sh

Every vmem command →

Session skills

Bundled in the plugin for Codex CLI and the ChatGPT desktop app. The Codex IDE extension uses the shared MCP setup, without the bundled skills.

Plugin source All four skills, explained →

Ready when you are

Start building your memory layer.

Connect Claude, Codex, Antigravity, or any MCP-aware client. Your memory follows you everywhere.

Get started →