← All posts
IntegrationsJuly 16, 2026·10 min read

Zsper is now live in Claude

Connect your Zsper brain to Claude in one click from the Claude directory. Write inside Claude grounded in your real opinions, stories, and voice — then sync the good stuff back so your brain sharpens for next time. The full guide: how to connect, every tool, and the loop that makes it compound.

Zsper is now a one-click connector in the Claude directory. That's the news. Here's why it's a bigger deal than "another integration went live."

You already live in Claude. You draft the cold email there, argue out the positioning, rough out the LinkedIn post between meetings. The problem was never Claude — it's a genuinely great place to think out loud. The problem is that Claude, like every chat window, starts each conversation as a polite stranger. It doesn't know the opinion you've defended across ten posts, the customer story you tell on every sales call, or the ₹ figure you quote when someone asks why Tier-2 towns are your best market. So you re-explain yourself, every session, forever.

Connecting Zsper closes that gap. Claude keeps being the room you think in. Zsper becomes the memory in the room — your real, scored, inspectable knowledge, available to Claude on demand — and the place the good stuff gets saved back to so it compounds.

TL;DR: Find Zsper in the Claude directory, connect with a sign-in (no token wrangling), and Claude can now read your Zsper brain — your opinions, stories, brand voice, and published articles — while you write. When you land on something worth keeping, Claude can sync it back to Zsper: a finished draft, a durable stance, a sharpened brand-voice signal. Everything read is workspace-scoped; everything written lands as a pending review you approve inside Zsper. Nothing publishes, nothing is trusted automatically. Write in Claude → sync to Zsper → your brain gets sharper → your next draft starts further ahead.

Connect it in about two minutes

The whole point of a directory listing is that there's nothing to configure. Here's the end-to-end path.

  1. Open the Zsper connector in Claude. Go to claude.ai/directory/connectors/zsper, or search "Zsper" in Claude's connector directory. It works in Claude on the web, the desktop app, and Claude Code.
  2. Connect with your Zsper account. Click connect and sign in the same way you sign into Zsper. Behind the scenes this runs a proper OAuth handshake — you're authorising Claude to reach one workspace's brain, and you can see and revoke that grant anytime. No copy-pasting a secret token out of a settings page (though a static token still works if you'd rather wire it up manually).
  3. Pick the workspace. The connection is scoped to a single workspace — one voice, one brain. If you run a workspace per client, you connect the one you want Claude working inside. It cannot reach across to the others.
  4. Start writing. That's it. In your next Claude conversation, ask it to check what you think before it drafts — "pull my stance on discounting from Zsper" — and it will.

If you manage several voices, the cleanest habit is one Claude project (or one connection) per Zsper workspace, so the fintech client's brain never bleeds into the D2C brand's.

What Claude can actually reach

Once connected, Claude has a real set of tools against your brain — not a vague "knowledge search," but specific, purposeful calls. It's worth knowing what's there, because the more precisely you ask, the better Claude uses them.

To find and read your knowledge:

  • search and fetch — search across your memory and articles, then open any result in full. These are the workhorses.
  • search_memory — hunt specifically through your durable knowledge: opinions, stories, frameworks, product facts, objections, personas, ranked by the brain's confidence.
  • search_articles — find your past drafts and published pieces by topic.
  • browse_memory — pull a whole category at once: "show me all my customer stories," "list the objections I've recorded."
  • get_related — from any one record, find its nearest neighbours by meaning, so Claude can gather the surrounding context before it writes.
  • list_recent_articles, list_drafts, list_topics — see what you've shipped, what's in progress, and how your topics are organised.

To write like you, not like a chatbot:

  • get_brand_voice — your audience, point of view, voice traits, tone do's and don'ts, signature phrases, banned words, and your measured prose rhythm. Claude should read this before drafting anything on your behalf.
  • check_voice_match — the one I'm quietly proudest of. It scores a draft against your actual published writing: a 0–1 voice-match number, the specific habits drifting from your style, learned AI-tell phrases, and a craft audit of robotic prose. Claude can draft, score, and revise until it stops sounding like a machine doing an impression of you.
  • get_brain_overview — a plain-language read on how well Zsper knows you and where it's thin. A good first call in a fresh chat.
  • get_writing_recommendations — what you should write next, computed from your own brain: pieces worth repurposing, high-confidence knowledge you haven't written up yet, open ideas — each with reasons.

To sync work back into Zsper (all review-gated):

  • save_draft — keep a finished article, newsletter, or LinkedIn post as a draft in your Zsper library.
  • save_idea — capture a content idea for later, filed at the ideation stage.
  • save_learning — remember a durable fact, opinion, story, or decision you stated in the chat.
  • update_draft — iterate across sessions on a draft you saved earlier; the previous version is snapshotted, never lost.
  • generate_brand_voice — fold substantial new brand signal into your voice profile so every future draft inherits it.

That's the surface. You don't have to memorise it — Claude picks the right tool. But knowing it exists changes how you ask.

What this looks like in your actual day

Abstract tool lists are useless. Here's the shape of it in real use.

Draft a sales email that already knows your objection-handling

You're replying to a prospect who's stalling on price. Instead of improvising, you tell Claude: "Draft a reply — pull my objection-handling on the 'too expensive' pushback from Zsper." Claude searches your memory, finds the exact framing you've used and refined, and writes the email in your voice, standing on language you already know converts. You're editing a reply that sounds like your best self on a good day, not one that reads like it came from a template farm.

Write a LinkedIn post from a stance you've already defended

Sunday idea, Monday post. You give Claude the rough thought; it calls get_brand_voice, pulls your relevant opinions and a customer story from search_memory, drafts the post, then runs check_voice_match on itself and fixes the lines that drift. Out comes a post that makes your argument with your Bengaluru client story — in Indian English, with rupees intact — not a generic hot take with a ₹ sign bolted on afterward.

Answer an investor question without contradicting yourself

An investor asks about your competitive position between meetings. You draft the answer in Claude, grounded in your recorded competitive insights and product facts. The value isn't speed — it's consistency. The position you give this investor is the same one you gave the last one, because it's retrieved from a single source of truth, not re-improvised and half-regretted each time.

Repurpose a long essay into a newsletter and a post

You published a 2,000-word article last month. In Claude: "Find my piece on pricing discipline in Zsper and spin it into a newsletter and a LinkedIn post." Claude fetches the article, reshapes it for each format's native rhythm, voice-checks both, and you sync the winners back as drafts. One argument, three formats, one voice.

In every case Claude isn't guessing what you think. It's retrieving it. That's the difference between an assistant that sounds plausible and one that sounds like you.

The loop that makes it compound

Here's the part that matters most, and the reason this is more than "read your notes in Claude." Writing in Claude and syncing to Zsper isn't a one-way export. It's a loop, and the loop is where the value accrues.

You write in Claude. Grounded in your brain, in your voice — the drafting is faster and lands closer to done.

You sync the good stuff to Zsper. The finished post goes back via save_draft. The sharp new opinion you articulated mid-conversation goes back via save_learning. The clearer positioning you finally nailed down goes back via generate_brand_voice. You approve each one inside Zsper — nothing is trusted on Claude's say-so.

Your brain gets sharper. Those approved records join your knowledge model. When you publish the draft from Zsper, the paragraphs you kept promote the records behind them; the ones you cut cost theirs a little confidence. Your memory doesn't just grow — it sharpens, because it learns what's actually you.

Your next draft starts further ahead. The next time you write — in Claude or in Zsper — you're standing on a brain that knows more and knows it more confidently. The fiftieth draft doesn't start where the first one did. That's the whole bet: knowledge is a by-product of the writing you were going to do anyway, and here it accrues even when the writing happens inside Claude.

Most integrations make a tool more convenient. This one makes your asset compound. Every conversation in Claude can leave your brain a little sharper than it found it — if you choose to save.

Safe by construction, both directions

I'll be blunt about the security model, because your name is on this work.

Reading is scoped and read-only-safe. A connection grants exactly one workspace — no cross-workspace reach. Retrieval hides suppressed and proposed records, the same guardrails as the app. Claude sees your confirmed knowledge, nothing you haven't stood behind.

Writing is never trusted automatically. This is the part people assume must have a catch. There isn't one. Everything Claude syncs back lands as pending: a learning arrives as a proposal that can't shape a single future draft until you approve it on your Brain page; a draft lands as an unpublished draft in your library. There is deliberately no publish tool, no delete tool, and no confirm tool on the connector. Claude cannot put words in your mouth, cannot ship anything, cannot quietly overwrite what you believe. Inbound text is sanitised on the way in. You are always the editor with the final say — the gate is wired into the architecture, not left to your discipline at 11pm.

And you can revoke the connection instantly from Claude or from Zsper's settings. The blast radius is bounded before you even do.

Why this fits an India-first workflow

Distribution in India runs on LinkedIn and newsletters, and Indian readers clock a US-written draft in a sentence. When Claude drafts off your Zsper brain, it inherits your Indian English by default — lakhs and crores, ₹ not $, a Nashik story that stays in Nashik instead of quietly relocating to San Francisco. You get Claude's fluency with your market's texture, instead of a brilliant draft that reads like it was written for someone else's country and find-and-replaced.

Get started

If you already use Zsper, connecting to Claude takes two minutes and changes where your brain can work: open the connector in the Claude directory and sign in.

If you're new, start with Zsper first — import a few pieces you've already published so your brain starts warm, then connect Claude and watch your assistant stop guessing what you think and start retrieving it.

Think. Remember. Write — now in Claude too, and syncing home every time.

  • Claude
  • MCP
  • Claude directory
  • write in Claude
  • AI memory

Start building your brain today.

Import what you’ve written, watch it learn, and ship a draft that sounds like you.

Start writing free