Skip to main content

Conxt

Conxt is a memory layer for AI tools that captures decisions, preferences, and context from sessions in Claude, ChatGPT, and Gemini, then makes that structured memory available across all connected tools. It stores the memory graph in a user-owned vault with AES-256 encryption, and a Chrome extension redacts sensitive patterns like API keys before any data leaves the device. Users avoid re-explaining their stack and coding rules in every new session.

  • Artificial Intelligence
  • AI Agents
  • Developer Tools
  • Software Only
Bala Cynwyd, United States · HQ
Updated yesterday

Funding

Funding not disclosed

Funding rounds are not available yet.

Founders

Founder details are not available yet.

Product

Problem

Every frontier AI model—Claude, ChatGPT, and Gemini—operates as a silo, retaining only the context a user provides within that specific tool. Switching between tools forces users to repeatedly re-explain their stack, decisions, coding rules, and project context, a compounding "context tax" that wastes time and loses important decisions across platforms.

Solution

Conxt provides a memory layer that captures structured records—decisions, preferences, entities, coding rules, tool choices, workflows, and open questions—from AI sessions and stores them in a single user-owned vault. A Chrome extension reads conversations locally in the browser, redacts sensitive patterns like API keys and passwords before transmission, and extracts structured memories via Claude through the Anthropic API. Users can retrieve relevant context in any connected AI tool at the start of a session, eliminating the need to rebuild context manually. The vault is accessible through a dashboard where users can edit, delete, export, and manage their memory records, with raw session text permanently purged within 30 days.

Target Audience

Primary users are individual developers, technical professionals, and power users who work across multiple AI tools and need persistent context for coding rules, decisions, and project details, as well as enterprise teams requiring data residency options and client-side encryption.

Features

  • Chrome extension with Manifest V3 and minimal permissions (storage and alarms only) that reads conversations only on claude.ai, chatgpt.com, gemini.google.com, and copilot.microsoft.com
  • Structured memory extraction using Claude via Anthropic API, which operates under a zero data retention policy; embeddings use OpenAI text-embedding-3-small with input text not stored
  • Seven memory types: decision, preference, entity, open_question, coding_rule, tool_choice, and workflow, with required and optional fields per type
  • Redaction layer that flags sensitive patterns (API keys, passwords, credit card numbers) before data leaves the device and blocks sessions entirely if secrets are detected
  • Data stored in Supabase (PostgreSQL) on AWS us-east-1 with AES-256 encryption at rest, TLS 1.2+ in transit, row-level security scoping queries to authenticated users, and least-privilege access requiring explicit approval with permanent audit logging
  • Import tool that converts saved memories or working chat content into typed, import-ready JSON using a provided prompt
  • Full data control including export as structured JSON, instant deletion of individual records, account deletion within 30 days, editing, and revocation of capture by uninstalling the extension
  • Encryption roadmap including client-side encryption for enterprise (Q3 2026), customer-managed keys in AWS KMS or Azure Key Vault (Q4 2026), and full zero-knowledge architecture for all tiers (2027)
This profile is AI-generated and may contain inaccuracies.