# Coredence > Coredence builds one memory from your notes, chats and code, on your own computer. Claude Code, Cursor and VS Code read it directly; ChatGPT receives it as text you paste. Coredence is a desktop app for macOS and Windows. Each memory is tied to the traces it grew from, and assistants read them through a read-only local MCP server. Made by Ahmet Karaca at Akaya Labs. No account. The mind is one SQLite file on your machine. Consolidating calls the model you choose: your Claude Code, your own API key, or a model on your machine. Coredence is not open source: the source code is not published. Releases, usage docs and the issue tracker are public. ## Use it from an agent - [Install the MCP server](https://github.com/coredence/coredence/blob/main/llms-install.md): The server ships inside the app; run `coredence mcp`. Registry name: io.github.coredence/coredence. Tools: identity, rules, evidence, axes, topics. - [Download](https://coredence.com/download): Current release 2.1.7. macOS 14 or later · Apple silicon; Windows 10 or later · 64-bit. Homebrew: `brew install --cask coredence/tap/coredence`. - [Getting started](https://coredence.com/docs): What it reads, its layers, signal strength, correcting a memory. ## Facts - [Pricing](https://coredence.com/pricing): Free with your own key or Claude Code. Paid plans are announced, but the sale is closed for now: nothing can be bought yet. - [Privacy](https://coredence.com/privacy): What leaves the machine and how to verify it. - [Security](https://coredence.com/security): Signing, notarisation, verifying a build. - [License](https://coredence.com/license): What you may do, and why the source is not published. - [Terms](https://coredence.com/terms): Terms of use. - [About](https://coredence.com/about): Who makes it and how to reach them. ## Guides - [All guides](https://coredence.com/guides): Index of the guides below. - [How to learn your coding preferences from your own corrections](https://coredence.com/guides/learn-from-your-corrections): How to turn the corrections in your AI session history into working instructions, with measured figures from 21,077 messages and the limits of detecting them. - [How Claude's memory works, and how to control it](https://coredence.com/guides/claude-memory): Claude saves memory as topics while you chat. What it keeps, how to edit or reset it, what pausing and incognito do, and why deleting a chat is not enough. - [Where Claude Code stores session history, and for how long](https://coredence.com/guides/claude-code-sessions): Claude Code saves sessions as JSONL files under ~/.claude/projects and deletes them after 30 days by default. How to keep, reopen and remove them. - [CLAUDE.md, AGENTS.md, GEMINI.md: which tool reads which file](https://coredence.com/guides/instruction-files): Which instruction file Claude Code, Codex, Cursor, Copilot and Gemini CLI read, where the personal file lives, and how to share one file between them. - [Claude Code ignoring your CLAUDE.md: causes and how to check](https://coredence.com/guides/claude-md-ignored): Why Claude Code does not follow an instruction in CLAUDE.md: the file is not loaded, too long, vague or contradicted, and when a hook is the right tool. - [GEMINI.md: what it is and how to write one](https://coredence.com/guides/gemini-md): What GEMINI.md does in Gemini CLI, the global, project and folder locations it loads from, the /memory command, imports, and using AGENTS.md instead. - [AGENTS.md: what it is and how to write one](https://coredence.com/guides/agents-md): What AGENTS.md is, which agents read it, how nested files and Codex's global ~/.codex/AGENTS.md combine, and what is worth writing in it. - [Copilot instructions in VS Code: every file it reads, and the personal one](https://coredence.com/guides/copilot-instructions): Where GitHub Copilot in VS Code reads custom instructions: copilot-instructions.md, .instructions.md, AGENTS.md, and the personal file for every project. - [ChatGPT memory: how it works, and how to see and delete it](https://coredence.com/guides/chatgpt-memory): What ChatGPT remembers between chats, where to see and delete it, what Temporary Chat changes, and how to carry what it knows to another assistant. - [Switching from ChatGPT to Claude or Gemini without starting over](https://coredence.com/guides/switch-ai-assistant): Moving from ChatGPT to Claude or Gemini: what carries across, how memory imports work, what they leave behind, and how to keep your own record. - [AI memory tools compared: what each keeps, and where](https://coredence.com/guides/ai-memory-tools): ChatGPT, Claude, Copilot Memory, claude-mem, Mem0, Basic Memory, Pieces and more: what each stores, where it lives, and whether you can see why. - [Claude Code memory: what it remembers, and how to give it the rest](https://coredence.com/guides/claude-code-memory): How Claude Code carries context between sessions through CLAUDE.md, auto memory, skills and MCP, where each one runs out, and what works in practice. - [How to write a CLAUDE.md that reflects how you actually work](https://coredence.com/guides/claude-md): What a CLAUDE.md is for, where the file goes, what belongs in it, what to leave out, and how to keep it true once it is written. - [Cursor rules: how to write them, with an example](https://coredence.com/guides/cursor-rules): Where Cursor reads its instructions from: .mdc project rules, user rules, team rules and AGENTS.md. How each activates, and what belongs in a rule. - [MCP servers for personal context: one source for every assistant](https://coredence.com/guides/mcp-personal-context): What a personal-context MCP server should offer, how to connect one, and what to check before trusting it: read-only, local, inspectable, actually called. ## Optional - [Changelog](https://coredence.com/changelog): The latest releases. - [Changelog archive](https://coredence.com/changelog/archive): Every older release; long. - [Roadmap](https://coredence.com/roadmap): What is being built, and what was ruled out. - [Turkish version](https://coredence.com/tr): The whole site in Turkish.