Stop re-explaining context · Free beta

Your AI forgets
what you told it yesterday.

The stack, decisions, and coding rules you build up in one AI tool are gone the moment you open another. Conxt captures them — and injects them into Cursor, Windsurf, Claude Code, Claude Desktop, and Codex through MCP.

Install Extension — Free →See how it works

Free during beta · No credit card · User-controlled capture

Memory capturing

claude.ai

Preference

Surgical fixes only, no unrequested refactors

claude.ai · just now

Decision

Use pdf-lib over pdfkit for Next.js PDF generation

claude.ai · just now

Entity

Conxt — cross-AI memory layer on Railway + Vercel

claude.ai · just now

Decision

Railway for FastAPI, Vercel for Next.js frontend

chatgpt.com · 2 min ago

Preference

Always test locally before pushing to production

claude.ai · 5 min ago

5 memories captured this sessionSee how it works →

Works withChatGPTClaudeGeminiMCPGrokImportPerplexityImportDeepSeekImportMistralImport

Claude, ChatGPT & Gemini sync live. Bring memory from any other tool into your vault with one-click JSON import.

Agent Memory

Your tools
don’t just read.
They write back.

Conxt is a read-write memory layer. Any MCP-connected tool can pull your context and record what it builds — attributed to the agent, tagged with the session, visible alongside your own decisions.

You decide in Claude, ChatGPT, or Gemini

Captured automatically — no action required

Your agent reads via get_context

Already knows your stack before the first line

Your agent writes back via add_memory

Attributed to the agent, tagged with the session

Next tool reads everything

Your decisions + agent work — nothing re-explained

CursorWindsurfClaude CodeClaude DesktopCodexAny MCP client

Connect your IDE →Now do it across your team →See all 14 tools →

Memory timeline

AllHumanAgent

Decisioncursor

session: run-0705

Reuse existing Stripe account — create new Products/Prices for Pro ($9/mo) and Team ($25/seat).

Coding Ruleclaude-code

session: run-0705

Webhook idempotency — store event_id, skip if already processed. Service-role client bypasses RLS.

Workflowdevin

session: run-0706

Deploy sequence: run migration → Railway → Vercel → register webhook endpoint in Stripe dashboard.

Tool Choice

claude

Use Stripe Checkout (hosted) over a custom payment form — all payment logic lives on the engine.

Decision

chatgpt

TypeScript strict mode always on — never use Tailwind color utilities, inline styles only.

Human decisions + agent work — one timeline, attributedSee all 14 tools →

For teams

Memory your whole team shares.

Every decision your team makes — in Slack, in Claude, in Cursor — captured once and available to everyone. No more “wait, what did we agree on?”

◈

Shared memory graph

Decisions made by anyone on the team are visible to everyone.

⊙

Slack capture

Decisions made in Slack are automatically extracted and stored.

→

Invite with one link

Share an invite link. Teammates join instantly — no admin setup.

Coming soon — $25/seat/moSee Conxt for teams →

KPZG LLC — team memory

3 members

Decision

slack #product

Launch on Product Hunt after ChatGPT capture is stable

Decision

slack #engineering

Railway for all Python services across KPZG projects

Entity

manual

KPZG LLC — Kalamazoo-based SaaS studio, two products

Question

slack #strategy

Whether to add onboarding flow before Product Hunt launch

The real upgrade

Stop repeating context.
Start every session already understood.

Conxt is not another chat app. It is the memory layer between the tools you already use — capturing what matters from one session and making it usable in the next.

No more copy-paste contextWorks across modelsBuilt for MCP workflows

Before Conxt

Explain your stack again

Repeat the same coding rules

Rebuild project context manually

Lose decisions across tools

After Conxt

Decisions captured once

Preferences follow you

Coding rules surface in your IDE

Every AI session starts with context

The problem today

Every frontier model is a silo.

Claude

claude.ai

Knows Only what you told Claude. Nothing about your other conversations.

ChatGPT

chatgpt.com

Knows Only what you told ChatGPT. Nothing about your other conversations.

Gemini

gemini.com

Knows Only what you told Gemini. Nothing about your other conversations.

Every time you switch tools — you re-explain your stack, your decisions, where you left off. That's the context tax. It compounds every single day.

Why ownership matters

Your memory graph. Not theirs.

Anthropic, OpenAI, and Google are private companies with shareholders and revenue targets. Their memory features live inside their systems — on their terms, subject to their pricing decisions, their policy changes, and their business model shifts. Conxt sits outside all of that. Your memory graph belongs to your account, not to the model you happened to be using when you figured something out.

Pricing changes

The free memory tier that worked for you today gets restricted tomorrow. Your accumulated context is still there — behind a paywall you did not plan for.

Model deprecated

The model you built months of context with gets sunset. The next version has no memory of your stack, your decisions, or where you left off.

Policy shifts

Memory features get rolled back, restructured, or limited in your region. Context you accumulated disappears without a clear migration path.

The Conxt answer

Your memory graph is stored in your account — structured, exportable, portable. The decisions you've made, the context you've built up across thousands of AI conversations, the preferences you've expressed — that's yours. Not subject to a pricing change. Not lost when you switch models. Not locked inside any one company's ecosystem. Connect it to Cursor, Windsurf, Claude Code, Claude Desktop, and Codex through MCP.

You own this

Export

Full JSON export at any time

Delete

Any record, any time, instantly

Portable

Cursor, Windsurf, Claude Code, Claude Desktop, Codex

Control

Capture and inject only when chosen

How it works

One graph.
Every tool.

Install once. Conxt silently connects your AI tools so they share what they know about you — decisions, preferences, context — across Claude, ChatGPT, Gemini, and every MCP-connected IDE.

01

Install the extension

Add Conxt to Chrome. It works immediately on Claude, ChatGPT, and Gemini — no configuration.

One click installWorks in the backgroundNo permissions beyond what's needed

02

Work normally

Use your AI tools exactly as you do today. Conxt listens silently and extracts structured memory — decisions, preferences, entities — from every session.

Auto-captureSmart extractionNo manual tagging required

03

Your memory loads automatically

Connect Cursor or Windsurf to your MCP endpoint once. From that point every new session starts with your decisions, preferences, and context already loaded — no button, no copy-paste.

MCP via Cursor & WindsurfPermanent API keyZero re-auth

Import from anywhere

Already have context in another AI? Bring it in one paste.

Run this in the tool you want to pull from. Paste in your saved memories, or run it at the end of a working chat — it converts what you give it into import-ready JSON, already tagged with the memory types Conxt understands. Drop the result into your vault and you're caught up. Run it in a blank chat and you'll get an empty array — there's nothing there yet to convert.

1Grab your memories

From your memory settings — or a working chat.

2Run the prompt

It converts them to typed, import-ready JSON.

3Drop it in your vault

Every memory lands, attributed to its source.

Prompt for your AI

Convert my memory into Conxt import JSON.

Use, in this order: (1) any memories I paste below,
(2) otherwise everything from this conversation.
If there's nothing to draw on, return an empty array — never invent anything.

Format:
{
  "memories": [
    { "type": "Decision", "text": "one specific memory", "source": "chatgpt.com" }
  ]
}

Every "type" must be exactly one of:
Decision, Preference, Coding Rule, Entity, Workflow, Tool Choice, Question

Return only the JSON — no explanation, no markdown fences.

MY MEMORIES:
(paste here, or leave blank to use this conversation)

Where to find them — ChatGPT: Settings → Personalization → Manage Memories. Claude & Gemini: your memory settings. Or run it at the end of any working chat to capture that conversation.

Supported types:DecisionPreferenceCoding RuleEntityWorkflowTool ChoiceQuestion

Built for real workflows

The more AI tools you use,
the more Conxt compounds.

The wedge is simple: people who already live across multiple AI systems should not lose their working memory every time they switch tabs.

Solo founders

Keep product context alive.

Carry positioning, pricing decisions, roadmap tradeoffs, launch plans, and customer research across every AI tool you use.

Product decisionsLaunch plansPositioning

Developers

Bring your coding rules with you.

Your stack choices, bug history, preferred patterns, and architectural constraints follow you into Cursor, Windsurf, Claude Code, and beyond.

Coding rulesStack contextArchitecture

AI power users

Move between models without resetting.

Use ChatGPT, Claude, Gemini, and MCP tools without rebuilding the same background context every time you open a new session.

ChatGPTClaudeGemini

IDE Integration

Your memory, live in your IDE.

Conxt exposes your memory graph as a remote MCP server. Point Cursor or Windsurf at one URL — your decisions, preferences, and context load automatically before every prompt.

CursorAdd to mcp.json — one URL, one key

WindsurfAdd to mcp_config.json — same key

Claude CodeOne terminal command

Claude DesktopAdd via Settings → Developer

CodexAdd to config — same key

Connect your IDE →

Your MCP endpoint

https://mcp.conxt.dev/mcp/Live

~/.cursor/mcp.jsonCursor ✓

{
  "mcpServers": {
    "conxt": {
      "type": "http",
      "url": "https://mcp.conxt.dev/mcp/",
      "headers": {
        "Authorization": "Bearer <your-api-key>"
      }
    }
  }
}

~/.codeium/windsurf/mcp_config.jsonWindsurf ✓

{
  "mcpServers": {
    "conxt": {
      "serverUrl": "https://mcp.conxt.dev/mcp/",
      "headers": {
        "Authorization": "Bearer <your-api-key>"
      }
    }
  }
}

One permanent API key works across every IDE. Generate it once at conxt.dev/connect-ide.

CAPTURE

SURFACE VIA MCP

Capture once · remember everywhere · nothing starts from zero

Features

Everything your memory needs.

Built for developers who live in multiple AI tools every day and are tired of re-explaining themselves.

Cross-AI memory

One memory graph across Claude, ChatGPT, Gemini, and Copilot. No silos, no context loss when you switch tools.

Smart extraction

Claude automatically extracts decisions, preferences, and entities from your conversations. Not raw text — structured intelligence.

MCP-native surfacing

Your context surfaces automatically at the start of every session in Cursor, Windsurf, Claude Code, Claude Desktop, and Codex.

Session grouping

Memories are organized by session, source, and type. Navigate your context graph without noise or clutter.

Privacy first

Raw session text is temporarily processed for extraction and purged within 30 days. Structured memory records persist until you delete them. Secrets and PII are blocked before storage.

You own your data

Export everything as JSON at any time. Delete individual records or your entire account instantly. No gotchas.

7 memory types captured automatically

DecisionsPreferencesEntitiesOpen questionsCoding rulesTool choicesWorkflows

Proof from the founder graph

This was built from a real memory problem.

Conxt came from measuring thousands of real AI-memory records and finding the same thing repeated in hundreds of subtly different ways. The product is designed for messy, paraphrased, cross-session context — not clean demo data.

4,189

memory records analyzed

60%+

duplicates found

604

records about one product

585

different phrasings

Pricing

Simple, honest pricing.

No surprise charges. Lock in beta pricing when you upgrade.

Free

forever

For individuals getting started with AI memory.

✓200 memory records

✓Manual capture (paste your session)

✓1 AI tool — read your own memories

—Auto-capture extension

—MCP IDE surfacing

Start free

Your AI tools
have amnesia.
Fix it.

Cursor, Windsurf, Claude Code, Claude Desktop, Codex. One install. Free during beta.

Get started free →

Founder

Priyadarshi Gavali

Founder, KPZG LLC

I built Conxt because I kept re-explaining myself to AI tools that should already know me. Every time I switched from Claude to ChatGPT to Gemini, I started from scratch — same stack, same preferences, same decisions, re-typed every session.

So I started capturing my own sessions to measure how bad it actually was. What I found was worse than expected: 4,189 memory records in my own graph, and over 60% were duplicates — 604 separate records describing the same product, written 585 different ways. Not identical enough for a hash to catch. Paraphrastic duplication that only semantic matching can resolve. I wrote a research paper about it.

I work in telecom retail during the day and build on nights and weekends. I'm finishing an MS in Computer Information Systems at Colorado State. Conxt is the infrastructure I wanted to exist — built solo, in public, measured on real data.

Read the preprint →hello@conxt.dev

Your AI forgetswhat you told it yesterday.

Your toolsdon’t just read.They write back.

Memory your whole team shares.

Stop repeating context.Start every session already understood.

Every frontier model is a silo.

Your memory graph. Not theirs.

Pricing changes

Model deprecated

Policy shifts

One graph.Every tool.

Install the extension

Work normally

Your memory loads automatically

Already have context in another AI? Bring it in one paste.

The more AI tools you use,the more Conxt compounds.

Keep product context alive.

Bring your coding rules with you.

Move between models without resetting.

Your memory, live in your IDE.

Everything your memory needs.

Cross-AI memory

Smart extraction

MCP-native surfacing

Session grouping

Privacy first

You own your data

This was built from a real memory problem.

Simple, honest pricing.

Your AI toolshave amnesia.Fix it.

Priyadarshi Gavali

Your AI forgets
what you told it yesterday.

Your tools
don’t just read.
They write back.

Stop repeating context.
Start every session already understood.

One graph.
Every tool.

The more AI tools you use,
the more Conxt compounds.

Your AI tools
have amnesia.
Fix it.