Portable memory layer for AI agents.
Memory that survives across sessions, models, and frameworks. Built on Solana for provable ownership.
| Package | Description | Tests |
|---|---|---|
@clude/brain |
Provider-agnostic memory engine (7-phase recall pipeline) | 25/25 ✅ |
@clude/cortex |
Portability, wallet identity, framework adapters (OpenAI, LangChain) | 35/35 ✅ |
@clude/cloud |
Supabase + Voyage AI + Venice providers | 21/21 ✅ |
@clude/local |
SQLite + gte-small local-first provider (zero config) | 15/15 ✅ |
@clude/bot |
X (Twitter) agent — the original Clude bot | — |
import { CludeEngine } from '@clude/brain'
import { SupabaseProvider, VoyageEmbeddings, VeniceLLM } from '@clude/cloud'
const engine = new CludeEngine({
storage: new SupabaseProvider({ url: SUPABASE_URL, serviceKey: SUPABASE_KEY }),
embeddings: new VoyageEmbeddings({ apiKey: VOYAGE_KEY }),
llm: new VeniceLLM({ apiKey: VENICE_KEY }),
scope: { owner_wallet: 'your-wallet' },
})
await engine.store({ content: 'Important fact', type: 'semantic' })
const memories = await engine.recall({ query: 'What do I know?' })import { CludeEngine } from '@clude/brain'
import { SQLiteProvider } from '@clude/local'
import { GteSmallEmbeddings } from '@clude/local'
const engine = new CludeEngine({
storage: new SQLiteProvider({ path: './memories.db' }),
embeddings: new GteSmallEmbeddings(),
})// All agents share the same engine → shared memory
const engine = new CludeEngine({ storage, embeddings, llm, scope })
// Agent A stores → Agent B recalls instantly
await engine.store({ content: 'Vector threshold = 0.25', source: 'researcher' })
const results = await engine.recall({ query: 'vector threshold' })See packages/cloud/examples/swarm-demo.ts for a full demo.
┌─────────────────────────────────────────────┐
│ Your Agent │
├─────────────────────────────────────────────┤
│ @clude/cortex (identity, portability, │
│ OpenAI/LangChain adapters) │
├─────────────────────────────────────────────┤
│ @clude/brain (7-phase recall, scoring, │
│ decay, entity graph) │
├──────────────┬──────────────────────────────┤
│ @clude/local │ @clude/cloud │
│ SQLite+gte │ Supabase+Voyage+Venice │
└──────────────┴──────────────────────────────┘
- Private — Venice AI for private inference, no data leaves your control
- Portable — Wallet-based identity, signed MemoryPacks, export/import anywhere
- Permissionless — No API keys needed for local mode, Solana for on-chain proofs
- Poly-model — Works with any LLM, any embedding model, any framework
- Persistent — Memories survive sessions, models, and infrastructure changes