
Codex vs Claude Code: which should you use, and how do you use both?
Codex vs Claude Code as of November 2026: plans, AGENTS.md vs CLAUDE.md, MCP config, memory, and how to run both on one project.
Blog
Field notes from Pavel on working with AI agents, keeping their context, and building Substrate.

Codex vs Claude Code as of November 2026: plans, AGENTS.md vs CLAUDE.md, MCP config, memory, and how to run both on one project.

An honest comparison by Substrate: when ChatGPT's built-in memory is enough, when it isn't, and how ChatGPT connects to Substrate as a custom MCP connector.

Claude's built-in memory vs Substrate, written by Substrate: when the built-in memory is enough, when it isn't, and a three-question test.

Substrate and Supermemory are different products. Where each one wins, plus pricing, MCP, Claude Code and open source, checked November 2026.

An honest comparison of Substrate and Zep by Substrate: pricing, open source, agents and sources, and when to choose each.

ChatGPT is the safer default, Claude is stronger for writing and long documents. The part nobody covers: carrying context between them.

An honest comparison of Substrate and Mem0 by Substrate: pricing, open source, agents and sources, a benchmark with caveats, and when to choose each.

What Claude remembers, where it is stored, how to edit it, and what it cannot carry to other tools, as of November 2026.

Claude Code and Cursor share no chat history. See what they do share, with file paths, and a ten-minute setup for teams running many agents.

How Claude Code memory works: CLAUDE.md levels and load order, auto memory, imports, what survives /compact, and how to keep it useful as you grow.

What an MCP memory server is, real Claude Code and Cursor config for the official one, and how it compares with four alternatives.