If you use ChatGPT for more than one type of work, you've probably developed a small collection of system prompts you paste in over and over. A detailed set of instructions for coding. A different tone for writing emails. Another for creative brainstorming. And every time you switch contexts, you manually clear the conversation and paste the new instructions.
It's repetitive, error-prone, and breaks your flow. Custom modes fix this by letting you save instruction profiles once and switch between them instantly — no copy-pasting, no retyping, no lost context.
What custom modes actually are
A custom mode is a saved instruction profile that tells ChatGPT exactly how to behave. Think of it as a persistent system prompt you can apply with a single click or keystroke.
Instead of starting every chat with "You are a senior developer who writes clean, well-documented code with TypeScript..." you save that as a mode called Developer. When you need it, you activate it. ChatGPT adopts that persona immediately.
You can create unlimited modes. Common examples include:
- Developer — strict, concise, code-first responses with examples
- Creative Writer — expressive, narrative-driven, open to ambiguity
- Editor — critical, precise, focused on grammar and clarity
- Researcher — thorough, skeptical, cites sources and asks follow-ups
- Support — patient, step-by-step, assumes minimal prior knowledge
- Click the mode name in the PromptsFlow sidepanel to apply it to the current chat.
- Use the Prompt Palette (`Shift+P`) to search modes by name and activate them without touching the mouse.
- Set a default mode so every new conversation starts with your preferred instruction profile automatically.
- Inconsistency — you forget parts of the prompt, or different versions exist in different places
- Friction — every context switch requires finding, copying, and pasting text
- Clutter — your first message in every chat is consumed by instructions instead of actual work
- No presets — there's no quick way to return to a previous configuration
- A "Skeptic" mode that challenges your assumptions
- A "Teacher" mode that explains complex topics simply
- A "Formatter" mode that converts rough notes into structured documents
Each mode stores a full system instruction string. The moment you switch, every new message in that conversation follows the new profile.
Creating your first mode
Setting up a mode takes under a minute:
1. Open the PromptsFlow sidepanel inside ChatGPT and navigate to Modes. 2. Click "Add Mode" and give it a name — something you'll recognize instantly like "Code Reviewer" or "Marketing Copy." 3. Write your instruction prompt in the text area. This is where you define the behavior, tone, formatting rules, and constraints. 4. Save it. The mode is now available across all your ChatGPT sessions.
Your instruction can be as simple or detailed as you want:
``` You are a senior frontend engineer. Respond with clean, modular code. Always explain your reasoning briefly before showing code. Use TypeScript and prefer functional components. Never skip error handling. ```
Or more open-ended:
``` You are a creative writing coach. Be encouraging but honest. Focus on voice, pacing, and emotional resonance. Ask questions rather than rewrite everything for me. ```
> Pro tip: Name your modes by outcome, not role. "Concise Email" is more actionable than "Assistant" because it tells you exactly when to use it.
Switching modes on the fly
The real power is speed. Once you've created a few modes, switching between them becomes effortless:
Switching mid-conversation is supported too. If you're halfway through a debugging session and want to pivot to documentation mode, activate the new mode and your next message follows the new instructions. Previous context stays intact — only the behavior rules change.
Why this beats manual system prompts
Manually pasting instructions seems fine until you do it ten times a day. Here's what breaks:
Custom modes solve all of this. They're stored, versioned, and accessible instantly. Your instructions are always exactly what you wrote. And because modes live in the sidepanel and palette, they never compete with your actual conversation content.
When modes make the biggest difference
Modes shine in two scenarios: context switching and collaboration.
If your workday bounces between tasks — writing code in the morning, reviewing copy in the afternoon, planning projects in between — modes eliminate the setup tax for each switch. One click and ChatGPT is configured for the new task.
For teams or shared workflows, modes also act as reusable templates. You can export your modes as JSON and share them with teammates, ensuring everyone uses the same instruction standards. A team of writers can share the "Brand Voice" mode. A dev team can standardize on the "Code Reviewer" profile.
Building a mode library that scales
Start small. Create 2–3 modes for your most common tasks. Use them for a week and refine the instructions based on the responses you get. Good modes are iterative — they improve as you learn what instructions produce the best output.
Over time, expand your library to cover edge cases:
---
_Custom modes are built into PromptsFlow — the browser extension that adds instruction profiles, a smart prompt library, folder organization, per-chat notes, and more to ChatGPT. [Try it free →](#)_