# Engram > Engram turns a folder of your own markdown notes into memory any AI can search, read, and write, from any device, with your computer closed. The notes stay plain files you own. You can run the whole stack on your own hardware. Engram is three interfaces over one backend: an Obsidian plugin that syncs an existing vault, a web app that browses and edits it, and a hosted MCP server at `mcp.engram.page` that serves it to AI clients over OAuth 2.1. Every documentation URL below has a raw markdown twin. Append `/index.md` to any docs path to fetch the source instead of parsing HTML, for example `https://engram.page/docs/concepts/open-knowledge-format/index.md`. The /compare pages use `https://engram.page/compare/.md`. Facts worth having straight, because they are commonly stated wrongly: - The hosted service is **not** end-to-end encrypted. Semantic search requires the server to read note text in order to embed it. Notes are encrypted at rest with per-user AES-256-GCM keys wrapped by AWS KMS. Self-hosting is the option for full data locality. - The backend is **source-available** under the PolyForm Small Business License, not open source. The Obsidian plugin is MIT. - Every tier, Free included, gets keyword, semantic and hybrid search. Free searches up to 2,000 indexed notes (the oldest 2,000 when a vault is larger); paid tiers search every note. The Free tier is **not** keyword-only. - Engram reads Open Knowledge Format v0.1 frontmatter natively but does not claim OKF conformance, and does not yet export bundles. ## Comparisons How Engram differs from the alternatives, including the cases where the alternative is the better pick. Each page has a raw markdown twin at `https://engram.page/compare/.md`. - [Engram compared](https://engram.page/compare): hub, with a where-each-wins table. - [Engram vs local Obsidian MCP servers](https://engram.page/compare/obsidian-mcp-plugins): A server inside your Obsidian process versus a hosted endpoint. The ceiling is architectural, not quality. - [Engram vs Obsidian Sync](https://engram.page/compare/obsidian-sync): Both sync your vault. Only one of them makes it searchable by an AI. - [Engram vs ChatGPT and Claude built-in memory](https://engram.page/compare/assistant-memory): Theirs is free and automatic. The difference is whether you can check a memory against a source. - [Engram vs Relay](https://engram.page/compare/relay): Relay ships live co-editing and we do not. Different problems on the same vault. - [Engram vs Notion and Mem hosted MCP](https://engram.page/compare/hosted-note-apps): Both have hosted MCP. The difference is whether your knowledge is a database or a folder. - [Obsidian Sync Alternatives](https://engram.page/compare/obsidian-sync-alternatives): Six ways to sync a vault without paying for Obsidian Sync, and the catch in each one. ## Start here What Engram is, and the philosophy behind it. - [Getting Started](https://engram.page/docs): What Engram is, how cloud and self-host differ, and how to get your first sync running in under five minutes. - [Why Engram](https://engram.page/docs/why-engram): The philosophy behind Engram. Why your notes belong to you, the honest tradeoff we made for AI-friendly memory, and why no vendor should lock your assistant in. ## Concepts How the parts work: search, sync, encryption, the Open Knowledge Format, and remote vs local MCP. - [Concepts](https://engram.page/docs/concepts): How Engram works under the hood. Semantic search, encryption, sync, supported file types. - [Encryption Model](https://engram.page/docs/concepts/encryption-model): Engram's encryption model: per-user data keys, AAD binding, which note fields are encrypted at rest and which are not. - [File Types](https://engram.page/docs/concepts/file-types): What Engram supports, how each type is handled, and what's not in scope. - [Open Knowledge Format (OKF)](https://engram.page/docs/concepts/open-knowledge-format): What Google's Open Knowledge Format is, how it works, and exactly which parts Engram reads today. - [Remote vs local MCP servers](https://engram.page/docs/concepts/remote-vs-local-mcp): Local MCP servers bind to localhost; remote ones are hosted HTTPS endpoints with OAuth. Why that decides whether your AI reaches notes from a phone. - [Semantic Search](https://engram.page/docs/concepts/semantic-search): How Engram searches your notes: chunking, embedding, vector and keyword indexing, and how results are ranked and retrieved. - [Sync Algorithm](https://engram.page/docs/concepts/sync-algorithm): How Engram detects, propagates, and reconciles changes across devices. ## MCP Connecting an AI client to a vault over the Model Context Protocol. - [About MCP](https://engram.page/docs/mcp): Model Context Protocol is the standard that lets AI assistants connect to tools like Engram. What it is, what you get, and what's under the hood. - [MCP Config](https://engram.page/docs/mcp/manual-config): The JSON config snippet for adding Engram to any MCP client. For users who prefer config files over the GUI flow. - [Scopes](https://engram.page/docs/mcp/scopes): What an MCP client can read and write once connected to Engram, which scopes it is granted, and how to revoke access. - [Tools](https://engram.page/docs/mcp/tools): Every tool Engram's MCP server exposes, what it does, and whether it reads, writes, or makes a sensitive write. - [MCP troubleshooting](https://engram.page/docs/mcp/troubleshooting): Fix MCP connection problems with Engram: Failed to connect, Authorization failed, expired tokens, and Claude finding no notes. ## AI integrations Per-client setup for Claude, ChatGPT, Cursor, and 13 others. - [AI Integrations](https://engram.page/docs/integrations): Connect Claude, Cursor, ChatGPT, and other AI assistants to your Engram vault. - [Antigravity](https://engram.page/docs/integrations/antigravity): Add Engram to Google Antigravity. One mcp_config.json covers the desktop app and the CLI; note that Antigravity requires serverUrl, not url. - [ChatGPT](https://engram.page/docs/integrations/chatgpt): Add Engram to ChatGPT as a Custom Connector. Developer Mode + OAuth, no config files. - [Claude Code](https://engram.page/docs/integrations/claude-code): Add Engram to Claude Code with one CLI command. OAuth-driven, no config files to hand-edit. - [Claude Desktop](https://engram.page/docs/integrations/claude-desktop): Add Engram as a Claude Desktop Connector. OAuth-driven, no JSON config files. - [Cline](https://engram.page/docs/integrations/cline): Add Engram to Cline's MCP Servers panel. API key in headers, no OAuth UI yet. - [Connect Engram to your AI](https://engram.page/docs/integrations/connect): One page, every client. Add Engram's remote MCP server to Claude, ChatGPT, Grok, Cursor, VS Code, Claude Code, or any MCP client. - [Continue](https://engram.page/docs/integrations/continue): Add Engram to Continue (continue.dev). Drop a YAML file, switch to Agent mode. - [Cursor](https://engram.page/docs/integrations/cursor): Add Engram to Cursor. Drop a config file or use the Settings UI; OAuth fires on the first tool call. - [Devin](https://engram.page/docs/integrations/devin): Add Engram to Devin's cloud agent. Register a custom MCP server in Settings, or one CLI command. - [Gemini](https://engram.page/docs/integrations/gemini): Connect Engram to the Gemini app as a custom Connected App for Spark. Paste the MCP URL, no credentials needed. Availability is heavily gated. - [GitHub Copilot](https://engram.page/docs/integrations/github-copilot): Add Engram to Copilot Chat in VS Code agent mode. Workspace `mcp.json` with OAuth or API key. - [Grok (xAI API)](https://engram.page/docs/integrations/grok): Wire Engram into Grok via xAI's API "Remote MCP Tools". For developers building on the xAI SDK. - [LobeChat](https://engram.page/docs/integrations/lobechat): Add Engram to LobeChat as a custom MCP plugin. Streamable HTTP + Bearer header. - [Mistral (Le Chat)](https://engram.page/docs/integrations/mistral): Add Engram to Mistral Le Chat as a custom MCP connector (OAuth, all plans), or wire it into La Plateforme agents via the Connectors API. - [Meta Muse](https://engram.page/docs/integrations/muse): Add Engram to Meta Muse as a custom connector. Remote streamable HTTP with OAuth. - [Open WebUI](https://engram.page/docs/integrations/open-webui): Add Engram to Open WebUI as a native Streamable HTTP MCP server. OAuth or API key. - [OpenCode](https://engram.page/docs/integrations/opencode): Add Engram to OpenCode with one JSON block. OAuth sign-in runs on the first call, then try it from any OpenCode session. - [Windsurf](https://engram.page/docs/integrations/windsurf): Add Engram to Windsurf's Cascade agent. Edit one JSON file or use the MCP panel. ## Obsidian plugin Installing and running Engram Vault Sync. - [Obsidian Plugin](https://engram.page/docs/obsidian): Install the Engram Vault Sync plugin, point it at a vault, and keep it in sync across devices. - [Conflicts](https://engram.page/docs/obsidian/conflicts): What triggers a sync conflict and how the resolution UI lets you choose what wins. - [First Sync](https://engram.page/docs/obsidian/first-sync): Sign in, choose Cloud or Self-Host, and complete the first-sync modal. - [Install](https://engram.page/docs/obsidian/install): Install the Engram Vault Sync plugin from Obsidian's community plugins directory. - [Mobile](https://engram.page/docs/obsidian/mobile): Engram on iOS and Android, what works, what's different, and known gotchas. - [Sync Center](https://engram.page/docs/obsidian/sync-center): The Sync Center panel covers current sync state, pending work, manual retry, and conflict resolution. - [Obsidian sync troubleshooting](https://engram.page/docs/obsidian/troubleshooting): What to do when sync stalls, sign-in fails, or a note refuses to push. ## Web app Browsing and editing in the browser. - [Web App](https://engram.page/docs/web-app): Browse, edit, and search your vault from any browser at app.engram.page. - [Browser](https://engram.page/docs/web-app/browser): Browse your vault in the Engram web app: folder layout, opening notes, the table of contents sidebar, and the mobile layout. - [Editor](https://engram.page/docs/web-app/editor): The Engram web editor: a Reading and Edit toggle with full Obsidian-style markdown rendering, wikilinks, callouts, math and diagrams. - [Onboarding](https://engram.page/docs/web-app/onboarding): The first-run setup at app.engram.page. Accept the Terms, then get straight to your vault. - [Search](https://engram.page/docs/web-app/search): Semantic search across your entire vault, with per-note exact match in the editor and full lexical search via Obsidian. ## Self-hosting Running the whole stack on your own hardware. Same code, same protocols. - [Self-Host](https://engram.page/docs/self-host): Run Engram on your own server with Docker Compose. Install, environment variables, encryption, backups. - [Architecture](https://engram.page/docs/self-host/architecture): What each piece of a self-hosted Engram deployment does, why it was chosen, and how the components connect on the network. - [Backup & Restore](https://engram.page/docs/self-host/backup-restore): What to back up in a self-hosted Engram deployment (Postgres, MinIO), how to restore it, and what a restore does not recover. - [Encryption Setup](https://engram.page/docs/self-host/encryption): How self-hosted Engram encrypts data at rest, how to back up and rotate the master key, and what the database looks like. - [Environment Variables](https://engram.page/docs/self-host/environment-variables): Complete environment variable reference for a self-hosted Engram deployment: required settings, optional tuning and defaults. - [Quickstart](https://engram.page/docs/self-host/quickstart): Docker Compose install in the fewest steps to a running Engram instance. Readable as docs; copy-paste-able as an AI prompt. - [Usage statistics](https://engram.page/docs/self-host/telemetry): What the Engram install census sends, where it goes, and how to turn it off. - [Self-host troubleshooting](https://engram.page/docs/self-host/troubleshooting): Common Engram self-host failures and how to fix them. Qdrant unreachable, encryption key errors, port conflicts, MinIO password, embedding queue stalls. - [Upgrade Path](https://engram.page/docs/self-host/upgrade): How to upgrade a self-hosted Engram install safely: the standard upgrade, reading release notes, rollback and major version bumps. - [Why Self-Host](https://engram.page/docs/self-host/why-self-host): When self-hosting Engram makes sense and what you trade off versus cloud. ## Billing Tiers, limits, upgrades, refunds. - [Billing](https://engram.page/docs/billing): Engram billing: the Free, Starter and Pro tiers, how upgrades and downgrades work, and how refunds and cancellation are handled. - [Refund & Cancel](https://engram.page/docs/billing/refund-cancel): How to cancel your Engram subscription, the refund policy, what happens to your data afterwards, and how to recover after auto-delete. - [Tiers](https://engram.page/docs/billing/tiers): Engram Free, Starter and Pro tiers compared: what each plan includes, the limits that apply, and which one fits your vault. - [Upgrade & Downgrade](https://engram.page/docs/billing/upgrade-downgrade): How Engram plan changes take effect: proration on upgrades, timing on downgrades, and what happens to your notes when limits change. ## Changelog What shipped, newest first. Synced from GitHub Releases. - [Changelog](https://engram.page/docs/changelog): Release notes for Engram are published on GitHub Releases. ## Product - [Everything Engram does](https://engram.page/features): the complete feature list, ending with an honest list of what is not built yet. - [Pricing](https://engram.page/pricing): Free, Starter $7/mo, Pro $14/mo. Per-tier limits. - [API reference](https://engram.page/docs/api): REST and WebSocket API, generated from the OpenAPI spec. - [Source](https://github.com/engram-app/engram): backend, source-available under PolyForm SBL 1.0.0.