22layout : post
33title : How to maximize Codex usage
44date : 2026-04-29 13:11:14
5+ updated : 2026-05-11 15:20:31
56excerpt : How to maximize your OpenAI Codex usage without going over the 5 hour window.
67categories : openai codex ai tui
78---
89
9- This post goes over how to maximize your [ OpenAI Codex usage] ( https://chatgpt.com/codex/settings/usage ) without going over the 5 hour window.
10+ This post goes over how to maximize your [ OpenAI Codex usage] ( https://chatgpt.com/codex/settings/usage ) so you don't go over the 5 hour window.
11+
12+ - [ Fast Mode] ( #fast-mode )
13+ - [ MCP] ( #mcp )
14+ - [ Model] ( #model )
1015
1116## Problem
1217
13- If you 're constantly hitting your 5 hour usage limit on Codex:
18+ You 're constantly hitting the 5 hour usage limit on Codex:
1419
1520> ⚠ Heads up, you have less than 25% of your 5h limit left. Run /status for a breakdown.
1621
1722> ■ You've hit your usage limit. To get more access now, send a request to your admin or try again at Apr 20th, 2026 4:20 AM.
1823
19- Then try disabling fast mode and/or changing your model.
24+ Try disabling fast mode and unused MCPs as well as changing your model.
2025
2126## Fast Mode
2227
@@ -39,6 +44,35 @@ If you have `service_tier = "flex"` in `~/.codex/config.toml`, then delete it or
3944service_tier = " flex"
4045```
4146
47+ ## MCP
48+
49+ Disable unused [ MCP] ( https://wikipedia.org/wiki/Model_Context_Protocol ) servers by editing your config:
50+
51+ ``` sh
52+ vim ~ /.codex/config.toml
53+ ```
54+
55+ Search for ` mcp_servers ` . For example:
56+
57+ ``` toml
58+ [mcp_servers .figma ]
59+ url = " https://mcp.figma.com/mcp"
60+ ```
61+
62+ To disable it, comment it out or set ` enabled = false ` :
63+
64+ ``` toml
65+ enabled = false
66+ ```
67+
68+ For example:
69+
70+ ``` toml
71+ [mcp_servers .figma ]
72+ url = " https://mcp.figma.com/mcp"
73+ enabled = false
74+ ```
75+
4276## Model
4377
4478If you're working on simpler coding tasks (e.g., writing tests), choose a smaller, faster, and more cost-efficient model like ` gpt-*-mini ` :
0 commit comments