Inside the 6 Sections of the open-context Preferences Editor
The open-context Preferences Editor has six sections — Communication Style, Technical
Profile, Work Context, Personal Context, Current Focus, and AI Behavior — plus a
standalone Custom Instructions field. Fill them out once at /preferences and
every vendor exporter (Claude, ChatGPT, Gemini) turns the same data into its own output
format, so you never re-describe yourself per provider.
Key takeaways
- The Preferences Editor has exactly 6 sections: Communication Style, Technical Profile, Work Context, Personal Context, Current Focus, and AI Behavior.
- A 7th field, Custom Instructions, sits outside the six sections and is appended verbatim to every vendor's exported preferences.
- Communication Style and AI Behavior map to the generated preferences document; Work Context, Personal Context, Technical Profile, and Current Focus map to the memory document.
- Every field is optional — empty fields are simply omitted from generated output, never filled with placeholder text.
- The schema is vendor-agnostic, so one pass through the editor exports to Claude, ChatGPT, and Gemini without re-entering anything.
Why one form feeds three AI vendors
Most AI assistants ask you to describe yourself from scratch — your role, your tone
preference, what you're working on — inside their own settings screen, in their own
words. If you use more than one assistant, you end up writing the same bio three times
in three slightly different dialects. open-context's
Preferences Editor fixes that by defining
a single vendor-agnostic UserPreferences schema. You fill it in once, and
each vendor
exporter translates that same structure into its own format — Claude gets prose
instructions, ChatGPT gets custom instructions text, Gemini gets Saved Info text.
Section 1: Communication Style
This section controls how responses should sound, not what they should contain. It has four fields and two checkboxes:
- Tone — formal, professional, neutral, friendly, or casual.
- Detail Level — concise, balanced, or thorough.
- Response Format — markdown, plain text, or structured.
- Language — a free-text language preference.
- Include code examples and Prefer step-by-step instructions — two independent checkboxes.
These fields map almost one-to-one onto sentences in the generated preferences document. Setting Tone to "friendly" and Detail Level to "thorough," for instance, produces a line like "I prefer warm and approachable communication with detailed, comprehensive explanations that cover edge cases" — no manual writing required.
Section 2: Technical Profile
Technical Profile captures your stack, not your personality. It holds an Experience Level selector (beginner through expert) and three tag-input fields — Programming Languages, Frameworks, and Tools — each accepting comma- or Enter-separated entries. Unlike Communication Style, this section doesn't feed the preferences document directly; it feeds the memory document, where your languages, frameworks, and tools are listed alongside your work context.
Section 3: Work Context
Three fields: Role, Industry, and a free-text Description. These exist to give an assistant enough professional context to calibrate its answers — a "Role" of "Backend Engineer" in the "Fintech" industry changes what counts as a reasonable default answer to a database question. None of this is considered sensitive by the Dashboard's privacy toggle in the technical sense, but Role and Industry are both blurred by that toggle since they're the two fields most likely to identify you if you're screen sharing.
Section 4: Personal Context
A Background textarea and an Interests tag input. This is the most clearly personally-identifying section in the editor by design — it's meant to hold things like hobbies or a one-line bio that a work-focused section shouldn't carry. It's also one of the fields the privacy toggle blurs first, for the same reason.
Section 5: Current Focus
Active Projects and Goals (both tag inputs) plus a Top of Mind textarea. This section is the most time-sensitive of the six — it's meant to be updated as your work changes, rather than filled in once and forgotten. When conversations are selected for export, the memory document also appends a list of your five most recently updated conversation titles directly under this section's content, so "what you're focused on" combines what you typed with what you've actually been talking to Claude about.
Section 6: AI Behavior
The last section controls how proactive the assistant should be, independent of tone. A Proactiveness selector (minimal, moderate, or proactive) sits above four checkboxes: ask follow-up questions when ambiguous, suggest alternative approaches, warn about potential risks, and assume context from previous messages. Like Communication Style, this section writes directly into the generated preferences document — setting Proactiveness to "proactive," for example, adds the line "Feel free to proactively suggest improvements or point out potential issues" verbatim.
The seventh field: Custom Instructions
Below the six sections sits a standalone Custom Instructions textarea, described in the UI as "free-form instructions that will be appended to every vendor export." It isn't counted among the six sections because it isn't structured data — it's raw text appended to the end of the generated preferences document, after everything the structured fields produce. Use it for anything the schema doesn't have a dedicated field for.
How the sections map to generated output
| Section | Feeds |
|---|---|
| Communication Style | Preferences document |
| AI Behavior | Preferences document |
| Custom Instructions | Preferences document (appended) |
| Work Context | Memory document |
| Personal Context | Memory document |
| Technical Profile | Memory document |
| Current Focus | Memory document |
For the Claude exporter specifically, that's a claude-preferences.md file
(built from Communication Style, AI Behavior, and Custom Instructions) and a
claude-memory.md file (built from Work Context, Personal Context, Technical
Profile, and Current Focus). See Claude Preferences vs Memory for what Claude
does differently with each of those two documents once pasted in.
Every field is optional
You don't need to fill out all six sections to get a usable export. Each exporter checks whether a field has content before writing a sentence about it — an empty Industry field just doesn't produce an industry sentence, rather than producing a sentence with a blank in it. That means you can fill in only Communication Style and still get a meaningful preferences document; the memory document will simply note "No personal context provided" for the sections you skipped.
Fill out your preferences once, export them everywhere.
Open the Preferences Editor →FAQ
Do I have to fill out every field in the Preferences Editor?
No. Every field is optional. Exporters skip empty fields when generating output — an empty Industry field, for example, simply isn't mentioned in the generated preferences document.
What's the difference between Work Context and Personal Context?
Work Context covers your role, industry, and a free-text description of what you do professionally. Personal Context covers your background and interests outside of work. Both feed the memory document, not the preferences document.
Where do Preferences Editor fields actually end up?
Communication Style and AI Behavior become sentences in a generated preferences document (claude-preferences.md for the Claude exporter). Work Context, Personal Context, Technical Profile, and Current Focus become a separate memory document (claude-memory.md). Custom Instructions is appended to every vendor's preferences export.
Does the Preferences Editor save automatically?
Yes. Every field change dispatches an UPDATE_PREFERENCES action to the app's React Context store immediately, so the preview on the Export page always reflects your latest edits.
Can I reuse the same preferences for ChatGPT and Gemini, not just Claude?
Yes. The Preferences Editor is vendor-agnostic — you fill it out once, and each exporter (Claude, ChatGPT, Gemini) translates the same underlying schema into its own output format.