Does Glen work with the Codex CLI and IDE integration?

Yes. Both support MCP, so you register Glen as a server in your Codex config and the agent calls it like any other tool to read and write shared memory.

Will each Codex run start cold?

No. Glen persists what Codex learns in your organization's shared store, so the next run, on any machine or by any teammate, starts with that accumulated context.

Shared memory for Codex

Codex is OpenAI's coding agent, run from the CLI or its IDE integration, and it supports MCP so you can plug in servers it calls while it works. It is strong at taking a task and driving it to a diff, but it has the memory of any fresh agent session: none. Codex does not recall the conventions it converged on yesterday, the reason a particular abstraction exists, or what a teammate's Codex already figured out about the same module. Every run re-learns the codebase. Glen, shared memory for AI agents, changes that by giving every Codex in your organization a single shared memory to read from and write to.

Add Glen as an MCP server in your Codex configuration and the agent gets one tool that both retrieves and records in a single round trip. When Codex picks up a task, it first pulls the context relevant to the files in play, the design decisions, the constraints, the do-not-repeat mistakes, so the diff it produces fits how your team actually builds. As it learns something worth keeping during the run, that observation goes back into the organization's shared store.

The org-scoped design is what turns this into leverage. Codex is often run repeatedly and across machines, and without shared memory each invocation pays the same cold-start tax. With Glen, the first Codex run to learn that an API client must be instantiated a certain way records it, and every later run, including a teammate's, inherits it. Because Glen works with any MCP client, the memory Codex builds is the same memory your Claude Code, Cursor, or VS Code agents read, so knowledge compounds across the whole toolchain rather than evaporating between sessions. Wiring it in is one MCP server entry authenticated over OAuth or an API key.

Add Glen to Codex

Add to ~/.codex/config.toml

~/.codex/config.toml

toml

[mcp_servers.glen]
url = "https://tryglen.com/api/mcp"

FAQ

Does Glen work with the Codex CLI and IDE integration?: Yes. Both support MCP, so you register Glen as a server in your Codex config and the agent calls it like any other tool to read and write shared memory.
Will each Codex run start cold?: No. Glen persists what Codex learns in your organization's shared store, so the next run, on any machine or by any teammate, starts with that accumulated context.