5 Claude Features You Didn't Know Existed

Most people use Claude the same way. Open a tab, type a question, read the answer, close the tab. Repeat.

That works. But it's also the equivalent of buying a professional camera and shooting everything on Auto.

Here are five features sitting right under the surface that most users never touch and probably should.

1. Projects: Stop Re-Explaining Yourself Every Session

Projects give Claude a persistent workspace. You load it once with your context a briefing doc, your role, tone preferences, standing instructions, and every conversation inside that Project starts from that foundation.

Most users skip this because the default Claude experience is stateless. Session ends, memory resets. Projects break that pattern entirely.

In practice: Create a Project for any ongoing work; a pitch deck, a client account, a research sprint. Paste in your reference material and set your instructions once. From then on, "draft slide 4" doesn't need three paragraphs of setup. Claude already has it.

Tip: Put your most important constraints in the Project instructions, not your first message. They'll apply to every conversation automatically.

2. Custom Style: Make Claude Write Like You

In Settings, you can upload samples of your own writing and Claude learns your voice your rhythm, sentence structure, level of formality. When you ask it to draft something, it sounds like you, not like a generic AI.

Most people assume AI writing always needs heavy editing. That's true when Claude is writing in its own voice. It's much less true when it's working from yours.

In practice: You run a newsletter. Upload three or four strong past issues. Give Claude a rough outline and get back a draft you're actually editing, not rewriting from scratch.

Tip: Upload writing you're proud of. The quality of your samples directly shapes the output.

3. Extended Thinking: Slow It Down on Hard Problems

Extended Thinking is a mode where Claude reasons through a problem step by step before answering in real time. It's slower, but it's also meaningfully better for anything genuinely complex.

Most people never enable it because speed is the default expectation. They get confident-sounding answers to hard questions and don't notice when something important got glossed over.

In practice: You're evaluating a major business decision, new market entry, restructuring a team. With Extended Thinking on, Claude surfaces its assumptions, flags uncertainty, and arrives at a more considered conclusion. Without it, you get a polished answer that might be missing something.

Tip: Save it for high-stakes questions. On simple tasks, it's just slower.

4. Prompt Caching: The API Feature Developers Skip Past

(This one's for developers building on the Claude API.)

Prompt caching lets you mark parts of your prompt; a long system prompt, a reference document, a knowledge base and store them server-side. Repeat API calls using the same cached content run faster and cost significantly less.

Most developers never configure this. They assume all API calls have roughly the same cost and latency. They don't.

In practice: Your app sends the same 10,000-token system prompt with every request. Without caching, you're paying full token costs each time. With caching enabled, repeat calls are cheaper and faster, which compounds quickly at any real volume.

Tip: See docs.anthropic.com for implementation. It's simpler than it sounds.

5. Voice Mode: Useful in the Hours You're Already Losing

Claude on mobile supports full voice conversations; you talk, it responds, you follow up. Not a dictation tool. An actual back-and-forth.

Most people write it off before trying it. Voice AI has a bad reputation (Siri, Alexa), and that reputation sticks. The difference is you're talking to the same Claude model you use in the browser.

In practice: You're commuting. You have a hard problem to work through; a tricky email, a decision you keep avoiding. Talk it out with Claude. By the time you arrive, you've done real thinking you'd otherwise have deferred.

Tip: Treat it like a conversation, not a command. The less you talk to it like a search engine, the better it gets.

Pick One and Try It This Week

Don't set up all five. That's how tools become noise.

Match the feature to your friction: ongoing work → Projects. Writing that needs your voice → Style. Hard decisions → Extended Thinking. High-volume API usage → Caching. Dead time in transit → Voice.

The gap between what Claude can do and how most people use it is wide.