Run Codex locally on Mac.

Step 1 · Launch via Ollama

One command to start.

$ ollama launch codex

Step 2 · Pre-flight memory check

Will your Mac fit it?

Codex pairs best with Qwen 2.5 Coder 32B at Q4_K_M (~19.0 GB). Add ~4–8 GB on top for the agent's working set, plus headroom for the rest of your dev stack. Recommended: 48 GB Mac (minimum 24 GB).

$ devpulse ai --before-load 19456 --auto-clean
before: Won't fit — 6.4 GB short
  - unloaded idle ollama model: qwen2.5:7b (4.2 GB)
  - killed 4 zombie procs (612 MB reclaimed)
after:  Fits comfortably — 3.1 GB headroom

# safe to launch
$ ollama launch codex

Why this matters: Fully agentic + large model = real OOM risk. Pre-flight every launch.

Step 3 · Babysit long sessions

Don't lose progress to OOM mid-task.

Agent runs that span hours hit memory pressure as context grows. devpulse babysit watches and auto-cleans without crashing the session.

# in another tmux pane
$ devpulse babysit --target-free-mb 8192 --json > codex.log &

# emits NDJSON tick / cleanup events you can checkpoint on:
{"event":"tick","tickNum":47,"availableForAIMB":7200,"pressure":"free<8192MB",...}
{"event":"cleanup","reasons":"free<8192MB","reclaimedMB":5400,...}

Tips for Codex

What we've learned running it.

• Codex tends to load files aggressively into context — pre-flight before pointing it at a large monorepo
• Pair with Qwen 2.5 Coder 32B for the best code quality at a memory budget that fits a 64 GB Mac
• If it crashes mid-edit, the cause is almost always the rest of your stack (Chrome, Docker) not the model. `devpulse processes -n 8` shows you who

Recommended models

Best local pairings for Codex.

Related agents