Live on Internet Computer Mainnet — 100% On-Chain

Give Your Agents
A Permanent Home.

Build agents with personas, tools, and memory. Deploy to any LLM. They return to Memory Camp when done — memory intact, ready to re-launch. No cloud. No servers. Encrypted on-chain.

See How It Works
Neural Memory Network — Live on ICP Mainnet
13
LLM providers
100%
On-chain storage
0
Servers, ever
Sessions remembered
The Agent Lifecycle

Build once. Deploy anywhere.
They always come home.

Total Recall is the home base every agent launches from and returns to.

🤖
Build
Define persona, attach tools, configure memory in Agent Builder
🚀
Launch
Generate a launch package with a secure session token — paste into any LLM
Mission
LLM reads memory, does the task, writes context continuously to Total Recall
🏕️
Memory Camp
Agent returns with a session summary — memory intact, ready to re-launch
HOW MEMORY GETS SAVED

Two modes. You choose.

LLMs have no built-in memory — Total Recall gives them two ways to keep it.

📝
Manual Mode
Agent decides what to save

The agent is instructed to save key facts, decisions, and context at intentional points — not everything, just what matters. Memory stays lean and clean.

✓  Save summaries, not noise
✓  Agent writes at natural checkpoints
✓  Best for long-running tasks & research
💡 How it’s enabled: Baked into the Launch Package system prompt. Add the 📝 Manual Save tool in Agent Builder for explicit checkpoints.
Automatic Mode
Every exchange is captured

Every message, output, and decision is continuously written to Total Recall throughout the session. Nothing slips through the cracks — even if the session ends unexpectedly.

✓  No gaps, no lost context
✓  Full session replay available
✓  Best for agents running live workflows
💡 How it’s enabled: Attach the ⬡ Always Store Context tool in Agent Builder — instructs the agent to write after every significant exchange.
⚠️ Requires function calling
Auto saving works by calling the Total Recall API mid-session. The LLM must support tool use / function calling to do this. Prompt-only models can’t make API calls — they can only output text.
🤖
Not all LLMs support automatic saving

Automatic saving requires the LLM to make real API calls during a session — that’s a function calling (tool use) capability. LLMs that only generate text can’t trigger saves on their own. They can still use Total Recall via Manual Mode with prompt-based instructions.

✅ Supports Auto Saving
GPT-4o / GPT-4 Turbo
Claude 3+ (Sonnet, Opus)
Gemini 1.5 Pro / Flash
Mistral Large
Cohere Command R+
Grok (with tools)
📝 Manual Mode only
Prompt-only models
Perplexity (no tool calls)
Basic Ollama / LM Studio
Models without tool use
Any chat-only interface
Fine-tuned instruction models
Both modes write to your on-chain canister. Mix them freely — most agents use automatic during the mission and manual for the final Memory Camp summary.
Everything You Need

The complete platform for
portable AI agents

🤖

Agent Builder

Create agents with custom personas, attach tools (webhooks, REST APIs, built-ins), configure memory keys, and manage multiple agents in one place.

Persona templates27 tool templatesMemory config
🌐

LLM Launch Pad

Pick your LLM and get a ready-to-paste launch package. No API key needed — use your existing ChatGPT, Claude, or Gemini account. Add automation keys only when running agents headlessly.

13 providersFree tier optionsKey testing
🏕️

Memory Camp

The agent’s home base. Browse and write memories, view session returns, and re-launch with one click — memory, persona, and tools intact.

Session returnsMemory explorerInline re-launch
⛓️

Workflow Chains

Chain multiple agents in sequence. Each agent reads the previous agent's output from on-chain storage — enabling complex multi-step AI pipelines that survive restarts.

Multi-agentPersistent stateAuto handoff
🗂️

Workspaces

Isolate memory environments by project. Every key is scoped to workspace:name/ — like git branches for agent memory. Switch contexts instantly.

Namespaced storageProject isolationSeed memory
🔐

Session Tokens

Your real API key stays private. Every launch generates a scoped sess_xxx token with a 24h TTL. Even if intercepted, damage is minimal and temporary.

Scoped accessAuto-expiryPIN protection
🔑

Secret Vault

Store seed phrases, API keys, and credentials encrypted with AES-256-GCM. Zero-knowledge — Total Recall never sees the plaintext. Your keys, your control.

AES-256-GCMClient-sideZero-knowledge
📁

File Storage

Store any file on-chain — images, PDFs, code, data. Chunked upload, permanent storage on the Internet Computer. No AWS, no Google, no single point of failure.

Any file typeChunked uploadPermanent storage
🤖

Canister Setup Agent

New to ICP? A built-in AI agent walks you through deploying your own canister — fund your wallet via QR code, create via NNS Dapp, see it live on-chain, and connect. No terminal required.

Guided setupAuto-connectStep by step
⛓️

100% On-Chain

Every byte lives on the Internet Computer Protocol. No cloud backend. No trusted server. The smart contract is the database — censorship-resistant by design.

ICP mainnetNo serversCensorship-resistant
Simple Pricing

Start free. Scale when ready.

Pay with card or ICP. ICP payments get 20% off.

Free
$0/mo
Your own ICP canister. Full sovereignty.
  • ✓ 100MB on-chain storage
  • ✓ 1 TR API key
  • ✓ 1,000 memory calls/day
  • ✓ Agent Builder (unlimited agents)
  • ✓ LLM Launch Pad (all 13 providers)
  • ✓ Workspaces + Workflows
  • ✓ Your own ICP canister
  • ✓ JS + Python SDK
  • 🤖 Built-in Canister Setup Agent
Agent
$39/mo
Built for production agent deployments.
  • ✓ 100GB on-chain storage
  • ✓ Unlimited TR API keys
  • ✓ 500,000 memory calls/day
  • ✓ Everything in Pro
  • ✓ Dedicated ICP canister
  • ✓ Team workspaces
  • ✓ SLA + dedicated support
All plans include: Agent Builder · LLM Launch Pad · Workspaces · Workflow Chains · Memory Camp · Session Tokens · Secret Vault · File Storage
Questions? Contact us →
Developer Tools

Connect in minutes.
Any language. Any LLM.

1
Install
npm install total-recall-sdk
2
Connect
new TotalRecallClient({ apiKey })
3
Store & recall
memory.store() / memory.get()
store(key, value, tags)
Save any string, object, or bytes
get(key)
Retrieve by key — null if not found
list_keys(prefix)
List keys, optionally filter by prefix
delete(key)
Remove a memory entry
stats()
Usage, tier, and call limits
handoff(id, name, msg)
Post session summary to Memory Camp

Ready to give your agents a home?

Free forever. No credit card. Connect with Internet Identity in seconds.

Built by Cleo 3 LLC · Powered by Internet Computer Protocol · Contact

Upload Files ?

☁️

Drag & drop files here

or

My Files ?

No files yet. Upload something!