Gemini Gems Alternatives for Reusable AI Instructions
Gemini Gems are not the only way to build a reusable AI assistant with standing instructions. This guide compares the main alternatives by what they let you write, attach and share.
Updated 5 min read
A Gemini Gem is a named AI assistant with standing instructions and optional reference files. If you want the same thing in a different AI app, every major assistant now has an equivalent: Claude has Projects, ChatGPT has Projects with custom instructions, and most coding tools let you set a persistent system prompt. Tools like fenbs let you store your tested instructions once and make them available across all of them, so you are not rewriting the same brief in every interface.
The right alternative depends on which AI app you use most, whether you need to share the assistant with a team, and whether you want the instructions to follow project context rather than just a general role. This guide covers the main options, where each fits and what to check before choosing.
What to look for in a Gems alternative
Before comparing tools, decide which capabilities matter for your workflow. Not every alternative offers all of them.
- Instruction length: how much text the tool lets you include in the standing brief.
- Knowledge files: whether you can attach documents the assistant draws on during the conversation.
- Team sharing: whether colleagues can use the same configured assistant without setting it up themselves.
- Cross-tool portability: whether your instructions work in other apps if you switch.
- MCP support: whether the tool can read from an outside record like fenbs at the start of each session.
Comparing the main alternatives
| Tool | Standing instructions | Knowledge files | Team sharing | Best fit for |
|---|---|---|---|---|
| Gemini Gems | Yes, in Gem setup | Yes | Yes, on Google Workspace accounts | Google Workspace users who want a quick configured assistant inside Gemini |
| Claude Projects | Yes, in Project instructions | Yes | Yes, on Team and Pro plans; check Anthropic's site for current plan details | Teams using Claude who want MCP support and deep instruction control |
| ChatGPT Projects | Yes, in Project custom instructions | Yes | Depends on plan; check OpenAI's help center | ChatGPT users who want per-project instructions after Custom GPTs retire |
| Cursor rules / .cursorrules | Yes, in project rule files | Via codebase indexing | Yes, file is checked into the repo | Developers who want instructions specific to a codebase |
| System prompt in the API | Yes, full control | Via retrieval if you build it | As far as your app shares it | Developers building AI features who control the full stack |
| fenbs | Skills and prompts stored outside any one AI app | No file storage; links to where files live | Yes, team workspaces with roles | Teams who want prompts and project context to travel across Claude, Cursor, Codex CLI, Gemini CLI, GitHub Copilot and Devin Desktop |
fenbs is not a direct replacement for a Gem. It does not run a conversation. It keeps the instructions, decisions and project context that every connected assistant should know before it starts, and it makes those available over MCP so a supported app can read them without you pasting anything in.
Claude Projects
Claude Projects are the closest structural equivalent to Gemini Gems. You write system instructions, upload knowledge files and start conversations inside the Project. Claude reads those instructions automatically. Claude Code Projects added shared memory on September 17, 2026, letting teammates see and build on a shared context.
Claude also supports MCP, which means you can connect fenbs and have Claude read your current project tasks, decisions and progress at the start of each session. That is something a standalone Gem cannot do. See the Claude Projects vs Gemini Gems comparison for a direct side-by-side.
Keep your instructions outside any one AI app
Store tested Gem instructions as Skills in fenbs. When you set up a Claude Project, a ChatGPT Project or a new Gem, the instruction text is already written and tested. Connected apps that support MCP can read Skills automatically; for Gemini and ChatGPT you paste them in. Create a free account to build your reusable instructions library.
Create a free accountFree for one person · No credit card · Works with Claude, ChatGPT, Cursor and more
ChatGPT Projects
ChatGPT Projects let you give a set of conversations a name, custom instructions and uploaded files. They work the same way as a Gem for personal use. OpenAI's Custom GPTs are being retired on December 11, 2026; check OpenAI's help center (opens in a new tab) for the current recommendation. For most everyday tasks, a Project with custom instructions is now the right option.
ChatGPT does not currently support MCP connections for outside data. That means project context, decisions and current task status must be typed in each session. For the Gemini equivalent in ChatGPT, the setup process and capabilities are similar.
Cursor and coding tool rules
For developers, Cursor and similar coding assistants support per-project instruction files (often called .cursorrules or similar). These are checked into the repository, so every team member gets the same instructions without any extra setup. The instructions are code-first: they describe conventions, the architecture and what the assistant should avoid.
Cursor supports MCP, so you can pair it with fenbs to add current task context alongside the static rules file. The instructions file sets the code style; fenbs tells the assistant which task is in progress and what decisions were made. That combination handles the parts that a static file cannot: current progress and evolving decisions.
Keeping instructions portable when you switch
Every tool stores its instructions inside its own interface. When you switch tools, you start from scratch unless you have kept the text somewhere outside. A straightforward habit is to write and test instructions in a document, then paste them into whatever tool you are using. The risk is that the copy in the tool drifts away from the original over time.
Storing instructions as Skills in fenbs gives you a single copy that any connected assistant can read. When you update an instruction, you update one place. For tools that do not support MCP, you still paste it in, but at least you are pasting from the current version.
How to do this with fenbs
- Create a free account and open your personal kanban board.
- Open Skills and prompts. Add a Skill for each standing instruction set you use regularly. Give each one a name matching the task it covers.
- Add the current project, its active tasks and any standing decisions to your board.
- Connect a supported AI app via Integrations. Ask it to read your Skills and the project context at the start of each session.
- For Gemini or ChatGPT sessions, copy the Skill text from fenbs and paste it into the Gem or Project instructions field.
fenbs connects to Claude, Claude Code, Cursor, Codex CLI, Gemini CLI, GitHub Copilot and Devin Desktop over MCP. It does not connect to Gemini Advanced or ChatGPT directly. The ai-prompts topic hub has more guides on building a reusable prompt library across tools.