Welcome to Total Recall
Your own private AI memory that feeds your IDEs and coding agents. Takes about 5 minutes to set up.
What does it do?
Total Recall gives your AI coding agents a persistent, sovereign memory. Your AI always knows who you are, what you're working on, and how you like code to be written. No cloud subscriptions. 100% local ownership.
- 🧠 Remembers across sessions — your preferences, projects, and decisions survive every new chat
- 🔗 Works with your AI tools — Claude Code, Codex, Gemini CLI, and any OpenAI-compatible app
- 🔮 Syncs with Obsidian — browse and explore your personal VFS memory files visually in a markdown note graph
- 🔒 100% private — runs entirely on your hardware under your control
How it works
Your memory lives in plain text Markdown files on your machine. Your AI agents read them automatically at the start of every conversation.
Configure sovereign VFS Brain Server
Set up your private VFS brain server. You can run the database server locally on this computer, or offload it to an always-on remote server/VM while keeping your local coding integrations active.
Regardless of where your brain server is hosted, the wizard will automatically configure this computer in Phase 4. This ensures your local IDEs, background log relays, and CLI coding agents (like Claude Code) can securely communicate with your memory files.
📍 Where to Host the Brain Server
Choose where to host your core plain-text VFS database and SQLite caches.
📂 Install Location — Select Brains to Provision
Choose where to install Total Recall memory brains. Select Global for universal identity rules and preferences, and/or individual project folders for project-specific knowledge.
🔑 Embedding API Keys (Semantic Search Only)
Total Recall routes all semantic search indexing tasks to lightweight embedding API calls (strictly for vector generation). All reasoning, codegen, and cognitive loops run 100% locally in your developer CLI coding agents (like Claude Code, Codex, or Antigravity) without any third-party routing. At least one key is required.
🌐 Public Access Sharing (Optional)
Configure secure remote access to your brain server.
*.trycloudflare.com HTTPS URL so you can connect coding agents and tools safely over the internet without configuring ports.Installing...
Total Recall is being provisioned on your server. This is a lightweight install and usually takes less than 60 seconds.
Auth & API Keys
Generate a Personal Access Token (PAT) to authenticate all your integrations.
What are PATs?
Every request to your brain must include a Bearer token. You can create multiple tokens with different scopes — one for Claude Code, one for Cursor, etc.
Generate your first key
📦 GitHub Automatic Backup optional
Securely back up your brain's VFS (your personal memory files, rules, and configurations) to a private GitHub repository automatically. We will create a private repository named total-recall-brain on your GitHub account, perform the initial push, and schedule a daily automatic update.
To set this up, create a Personal Access Token (PAT) with repo scope.
repo scope.
🔍 Web Search for Automatic Research optional
Total Recall runs in the background and automatically looks things up on the web — finding relevant articles, documentation, and facts related to what you're working on, and saving them to your memory.
You don't need a key to get started. Without one, it uses DuckDuckGo and Wikipedia (both free, no sign-up). Adding any one of the providers below gives it full web search results. Pick whichever you prefer — they're all roughly equivalent in quality.
Designed specifically for AI research — returns clean, readable text from each page instead of just a link and a short description. Saves time. 1,000 free searches/month.
Independent web search (not Google). Returns a list of URLs and short descriptions for each result. Solid general-purpose choice.
Finds pages by meaning rather than exact keywords — good at surfacing recent articles and blog posts that are thematically relevant, not just keyword matches.
Returns Google Search results. The free credits are a one-time trial rather than a monthly allowance, so it's best used as a last fallback or for occasional use.
~/.agent/config/research.yml.
Connect Your Tools
Select every IDE and agent you use. Total Recall will wire them up automatically — no manual config needed.
Which IDEs and coding agents do you use?
Select all that apply. Total Recall's relay will watch their session logs and compile your memory automatically.
🛠️ CLI Agents Setup & Installer
Install globally and initialize your selected CLI coding agents on the target computer in one click.
@anthropic-ai/claude-code)npm link)@openai/codex)🔄 Relay Daemon
The relay watches your IDE session files and ships them to your brain every 60 seconds automatically — no manual steps needed.
💾 Backup & Sync
Browse your memory graph visually or download a full backup of your brain.
📥 Import Existing Rules & Skills
Already have AGENTS.md, .cursorrules, CLAUDE.md, or modular Skills (under .agent/skills)? Scan and import them so Total Recall compiles them into every session.
API Reference
Full endpoint list for your brain. All requests require Authorization: Bearer <PAT>.
OpenAI-compatible
| Method | Path | Description |
|---|---|---|
| POST | /v1/chat/completions | Chat completions (streaming supported) |
| GET | /v1/models | List available models |
Memory
| Method | Path | Description |
|---|---|---|
| GET | /memory | List all memory files |
| GET | /memory/:path | Read a memory file |
| POST | /memory/:path | Create / overwrite memory file |
| PUT | /memory/:path | Append to memory file |
| DELETE | /memory/:path | Delete memory file |
| GET | /memory/search?q=… | Search memory |
| POST | /vault/compile | Recompile the vault surface |
| GET | /vault/status | Last compile status |
Keys & Sessions
| Method | Path | Description |
|---|---|---|
| GET | /api/keys | List keys |
| POST | /api/keys | Create key |
| DELETE | /api/keys/:id | Revoke key |
| GET | /api/sessions | List sessions |
| DELETE | /api/sessions/:id | Delete session |
Sandbox & Discovery
| Method | Path | Description |
|---|---|---|
| POST | /sandbox | Execute code in sandbox |
| GET | /health | Health check (no auth) |
| GET | /api | Full API reference JSON |
| GET | /.well-known/total-recall.json | Discovery manifest |
Example curl request
🎉 Your Brain is Live
Total Recall is installed and ready. Here are your URLs.
🔑 Dashboard Sign In Credentials
Use the credentials below to log into your settings dashboard (/dashboard). For security, you will be forced to change this temporary password upon your first access.
What's next?
- 💾 Start chatting — open the dashboard or point Claude Code at your brain
- 🧠 Build your memory — run
npx total-recall ingestto import your chat history - 🔄 Keep it synced — set up
npx total-recall sync --watchfor continuous ingest - 📚 Docs — github.com/gregiteen/total-recall
🔧 Manual Endpoint Override
If your secure tunnel URL or host endpoint is stale (e.g. http://ssh1.vast.ai:10120), or you have a custom Cloudflare tunnel URL, you can manually override your brain server endpoints below.
⚙️ Settings
Manage your search providers, usage limits, and other preferences.
🔍 Web Search Providers
The background research engine tries these in order — Tavily → Brave → Exa → Serper — and falls back to DuckDuckGo (free) when none are set or the daily limit is reached.
Returns full page text, not just links. 1,000 free searches/month. tavily.com
🔑 Total Recall API Key
This is the primary Personal Access Token (PAT) used by extensions and IDEs to securely connect to your memory vault.
🔒 Saved securely in your local configuration. You can copy it directly at any time (even when hidden) or click Show to reveal it. If you need a new key, you can regenerate one at Phase 3 (API Key).
🔑 Remote Dashboard Password
Reset or change the administrator password of your active local or remote brain instance.
⚠️ Danger Zone
Reset all configurations to start fresh. This will clear the wizard's memory from this browser and the server.
This clears both local storage and the server-side configuration file (`~/.agent/config/wizard-config.json`). If you deployed to a Cloud VM/VPS, remember to destroy or stop it directly in your cloud provider's console to avoid ongoing charges.