CortexDB memory for a Node agent (OpenClaw)

Wire CortexDB in as the memory layer for a Node.js agent. CortexDB is a pure-Go, single-file database; the Node agent talks to it over gRPC via the cortexdb-client package. Beyond vector/lexical recall, it gives the agent a real knowledge graph (RDF + SPARQL) — the thing most agent-memory layers lack. It fits OpenClaw's local-first philosophy: one binary, one SQLite file, no separate service to stand up.

When to use this

• The agent should remember user facts/preferences across turns or sessions.
• The agent should recall memories by meaning, not exact match.
• The agent needs entities + relations and multi-hop questions ("who, among the people Alice knows, works on X").
• You're integrating with OpenClaw (or any Node agent/skill).

Step 1 — Run the sidecar (once)

# install the binary (or download a prebuilt release)
go install github.com/liliang-cn/cortexdb/v2/cmd/cortexdb-grpc@latest

# lexical mode — zero config, no API key:
CORTEXDB_PATH=agent.db CORTEXDB_GRPC_TOKEN=s3cret cortexdb-grpc
# → listening on 127.0.0.1:47821

Enable vector/semantic recall by pointing it at any OpenAI-compatible embeddings endpoint (e.g. a local Ollama):

OPENAI_BASE_URL=http://localhost:11434/v1 \
CORTEXDB_EMBED_MODEL=embeddinggemma CORTEXDB_EMBED_DIM=768 \
CORTEXDB_PATH=agent.db CORTEXDB_GRPC_TOKEN=s3cret cortexdb-grpc

Step 2 — Install the client

npm install cortexdb-client

Step 3 — The two core moves: remember + recall

const { CortexClient } = require('cortexdb-client');

const client = CortexClient.connect('127.0.0.1:47821', { token: 's3cret' });

// remember a fact about the user (scoped per user)
await client.memory.SaveMemory({
  memoryId: 'pref-coffee', userId: 'alice', scope: 'user',
  content: 'Alice prefers dark roast coffee and runs OpenClaw locally.',
});

// later turn / next session: recall by meaning
const hits = await client.memory.SearchMemory({
  query: 'what does the user like to drink?',
  userId: 'alice', scope: 'user', topK: 3,
});
for (const h of hits.results) console.log(h.memory.content, h.score);

Every RPC is a promise; request fields are camelCase. Memory scopes isolate data: scope: 'user' (per userId), scope: 'session' (per sessionId), or scope: 'global'.

Step 4 — Knowledge instead of plain memory (RAG)

await client.knowledge.SaveKnowledge({
  knowledgeId: 'doc-1', title: 'Project brief',
  content: 'The user is building an autonomous agent in TypeScript.',
});
const res = await client.knowledge.SearchKnowledge({
  query: 'what is the user building?', topK: 3,
});

Step 5 — The differentiator: a knowledge graph

const iri = (v) => ({ kind: 'iri', value: v });
await client.graph.UpsertNamespace({ prefix: 'ex', uri: 'https://example.com/' });
await client.graph.UpsertKnowledgeGraph({ triples: [
  { subject: iri('ex:alice'), predicate: iri('ex:knows'), object: iri('ex:bob') },
] });
const ans = await client.graph.QuerySparql({
  query: 'SELECT ?o WHERE { \x3Chttps://example.com/alice> \x3Chttps://example.com/knows> ?o . }',
});
console.log(ans.result.count, 'result(s)');

Expose CortexDB as OpenClaw tools

OpenClaw skills teach the agent how and when to call tools. A ready-to-use helper module wraps the calls above into remember, recall, saveKnowledge, searchKnowledge, relate, and askGraph:

• scripts/memory-tools.js — import these and register them as OpenClaw tools, or call them directly from a custom skill/plugin.

In your skill's instructions, tell the agent: to remember a durable fact about the user, call remember(text); to recall, call recall(query); to record a relationship, call relate(subject, predicate, object); to answer a structured "who/what is related to X" question, call askGraph(sparql).

Install this skill into OpenClaw

OpenClaw follows the agentskills.io spec and discovers skills under \x3Cworkspace>/skills, \x3Cworkspace>/.agents/skills, ~/.agents/skills, and ~/.openclaw/skills.

# from this repo (local directory):
openclaw skills install ./skills/cortexdb-memory-openclaw --as cortexdb-memory

# or from git / ClawHub:
openclaw skills install git:liliang-cn/cortexdb@main --global

--global installs to ~/.openclaw/skills. The skill becomes eligible automatically once the cortexdb-grpc binary is present and a sidecar is running.

Sub-clients (full surface)

client.knowledge, client.memory, client.graph (RDF/SPARQL/SHACL/inference/ ontology), client.graphrag, client.tools (generic dispatch, same shape as MCP), client.admin. Each RPC is available in both PascalCase (SaveMemory) and camelCase (saveMemory). Auth is a bearer token; pass { token } to connect.

Notes & gotchas

• Zero-key default: without an embedder the sidecar uses lexical retrieval — good enough to start, no credentials needed.
• One file, one process: the sidecar owns one SQLite file. Isolate multiple users via memory scopes (above), not multiple files.
• Plaintext localhost: the bearer token rides plain gRPC; fine on localhost, add TLS / a reverse proxy for cross-machine use.
• No build step: the npm client loads the proto contract at runtime.
• Package and docs: https://www.npmjs.com/package/cortexdb-client · https://github.com/liliang-cn/cortexdb