Short answer: The simple baseline is one preferences document that you paste into each assistant's custom instructions. If you want the assistants to read the same file and update it as your preferences change, keep it in Fulcra and connect each assistant to it, so there is one copy instead of three that drift apart.
The problem
Claude, ChatGPT and Gemini each keep their own custom instructions and memory. Tell one you prefer metric units or short answers and the others never hear about it. Keeping them aligned means editing three settings pages by hand, and they drift as soon as one assistant learns something new.
When this is a good fit
- You use two or more assistants or agents regularly and change your preferences often enough that manual copying gets stale.
- You want assistants to record new preferences themselves, in one place you can read and edit.
- You want version history so you can see when a preference changed and roll it back.
When it isn't
- Your preferences are short and stable. One doc pasted into each tool is simpler and free.
- You only use one assistant. Its own custom instructions and memory are enough.
- The assistant you use can't connect to external tools. Fulcra's Gemini page says the consumer Gemini app can't add custom connectors yet; there, pasting is the only route.
Other options
- One preference doc pasted into each tool. Write it once, paste it into each assistant's custom instructions; update by hand.
- A notes file you own (Markdown, Obsidian or similar) as the source of truth, copied into each tool when it changes.
- Other MCP memory servers such as mem0, Supermemory or Basic Memory. Compare them on hosting, who owns the data and export.
- Connect each assistant. Gemini CLI:
gemini mcp add --transport http fulcra https://mcp.fulcradynamics.com/mcp, then sign in with /mcp auth fulcra.
- Seed the file. Tell one assistant: "Write my preferences to
/agents/preferences.md in Fulcra," and paste your existing doc.
- In each assistant's custom instructions, add: "Read my preferences from
/agents/preferences.md in Fulcra at the start of a conversation, and update the file when I state a new lasting preference."
Agents with a shell (Claude Code, Codex, Gemini CLI) can use Fulcra's fulcra-prefs skill instead, which keeps preferences as records in a "User Preference" data type rather than a file:
npx skills add fulcradynamics/agent-skills --skill fulcra-prefs
FAQ
- Is it free? There's a free tier with 5 GB, no credit card.
- Who can see my preferences? Only agents you connect. Data is encrypted at rest and in transit and never sold; you can export or delete it any time. Agent access is kept in check. Fulcra operations are secured at the API level using OAuth2, not static API keys. All access operations are logged so you can trace it.
- What if two assistants update the file at once? One write replaces the other, but every version is kept and can be restored.
Written by Fulcra. We make one of the tools recommended on this page.
.png){kind=logo}
