OpenAI's Codex v0.134 adds conversation search, goal mode, and Appshots
The most important feature in OpenAI Codex's latest update isn't a new model or a flashy agent — it's the ability to find past conversations. That seemingly small addition reveals how the battle for developer AI assistants is shifting from raw capability to everyday usability.
OpenAI's Codex AI coding assistant v0.134 introduces local conversation search with case-insensitive matching and previews, improved MCP setup with per-server environment targeting and OAuth options, and concurrent read-only tools when annotated readOnly. It also delivers more reliable tool schemas, richer extension and hook context, and fixes for Windows TUI rendering and remote session reliability, along with clearer docs and packaging updates.
- Conversation search transforms Codex from a stateless assistant into a persistent personal coding memory, reducing context loss across sessions.
- Per‑server MCP environment variables and OAuth enable safer multi‑model workflows, a capability competitors like Copilot lack.
- Windows TUI fixes expand Codex's addressable market, but pricing uncertainty (API consumption vs. flat‑rate) remains a barrier for heavy users.
Why It Matters
AI coding assistants are moving from code generation to comprehensive workflow tools; search is a prerequisite for that evolution.