Your agent,
ready to go

Wire up agentic-core + agentic-memory into a living agent.
Multi-session. Knowledge-aware. Lifecycle built in.

~3KB
gzip
0
dependencies
1
function
GitHub See Usage ↓
$ npm i agentic-claw agentic-core agentic-memory copy

Five lines to an agent

No framework. No boilerplate. Just createClaw and go.

app.js
const { createClaw } = require('agentic-claw') const claw = createClaw({ apiKey: 'sk-...', }) // Chat with memory await claw.chat('What is quantum computing?') await claw.chat('Tell me more') // remembers // Knowledge await claw.learn('docs', longDocument) await claw.chat('Summarize the doc') // Multi-session const alice = claw.session('alice') const bob = claw.session('bob') // Lifecycle claw.heartbeat(() => { ... }, 60000) claw.schedule('5m', () => { ... }) claw.on('token', d => process.stdout.write(d.text))

How it works

claw wires the building blocks. You just call chat().

your app
CLI / Web / Paw / Your App
↓
runtime
🦀 agentic-claw — chat · session · heartbeat
↓
building blocks
⚡ core — LLM · tools · schema
🧠 memory — context · knowledge

What's inside

🧠 Auto Memory

Every conversation automatically stored, trimmed, and managed. No manual history threading. Just chat.

🔍 Knowledge Recall

Learn documents once, recall relevant context on every chat. Local TF-IDF or OpenAI embeddings.

👥 Multi-Session

Isolated conversations, shared knowledge. Each session has its own history but taps the same knowledge base.

💓 Heartbeat + Schedule

Periodic callbacks and scheduled tasks. Your agent stays alive and can do work between conversations.

🔌 Streaming

Token-by-token streaming with event callbacks. Proxy support for GFW and CORS. Works everywhere.

📦 Zero Dependencies

Peers: agentic-core + agentic-memory. No transitive deps. No build step. Works in Node and browser.

How it compares

LangChain
Vercel AI SDK
agentic-claw
Install size
~80MB
~20MB
~30KB
Dependencies
200+
40+
2 peers
Memory
Extra package
Manual
Built in
Knowledge/RAG
Extra setup
Manual
learn() + recall()
CLI
None
None
claw-cli
Time to first agent
~1 hour
~30 min
~2 min