v3.0 · Open Source

Programmable
Autonomous Workers

The operating system for autonomous AI agents. Multi-channel, multi-agent, safety-first. Every action follows a strict six-stage pipeline — the LLM reasons, the system executes.

Get Started View Source ↗
you › "Scrape competitor pricing and save to a spreadsheet"
paw › 📋 Plan: 3 steps · navigate → extract → file_write
paw › ✅ Step 1: Opened target page
paw › ✅ Step 2: Extracted 47 products
paw › ✅ Done — saved to data/pricing.csv · 1.2s
30+
Built-in Tools
11
Channels
79
Tests
6
Pipeline Stages
6
AI Providers
How It Works

Six-Stage Safety Pipeline

Every action — without exception — passes through six validated stages. The LLM generates plans. The system executes them. They never cross.

01
🛡️
Intent

Sanitize input. Rate limit. Detect prompt injection across 15+ patterns.

02
🧠
Plan

LLM generates strict JSON plan. No execution, no side effects, reasoning only.

03
✅
Validate

Schema check. Safety policy. Blockchain simulation. Risk scoring.

04
⚡
Execute

Tools, APIs, browser, blockchain. Sandboxed and scoped. No raw LLM access.

05
🔍
Verify

Confirm results. Self-heal failures. Diagnose → fix → retry → escalate.

06
📋
Log

Full Clawtrace audit trail. Auto-redacted secrets. JSONL structured output.

Capabilities

Everything You Need,
Nothing You Don't

Production-grade components that work together. Extend with skills, MCP servers, or custom tools.

🤖

Autonomous + Supervised

Per-user mode toggle. Low-risk actions auto-execute, critical actions always confirm. Safety pipeline runs in both modes.

🌐

Browser Automation

Puppeteer-based control. Navigate, click, type, extract, screenshot. Full headless browser in the agent pipeline.

🔗

Multi-Agent Orchestration

Agent registry with capability routing. Delegate tasks, coordinate work, enforce depth limits across sub-agents.

🧠

Vector Memory

Persistent semantic memory. TF-IDF embeddings with cosine similarity search. Scoped by session, user, or global.

⛓️

Solana Native

Transfers, balance checks, SPL tokens, transaction simulation. On-chain agent registry with PDA-derived identity.

🔌

MCP Protocol

Connect to any Model Context Protocol server. Discover and invoke external tools without writing code.

📊

DAG Workflows

Visual workflow engine. Triggers, conditions, parallel branches, error handling. No external orchestrator needed.

🔐

Key Management

AES-256-GCM encryption at rest. Ed25519 signing. Keys zeroed after use. Never logged, never leaked.

🔄

Self-Healing

Intelligent failure recovery. Diagnose root cause, apply fix strategy, retry intelligently, escalate only when needed.

Connectivity

Works Everywhere You Do

Eleven channels. One brain. All sharing the same tools, memory, and safety pipeline. Install only what you need.

Telegram Discord Slack WhatsApp Email SMS WebChat Webhooks LINE Reddit Matrix

Telegram via Telegraf · Discord via discord.js · Slack via Bolt · WhatsApp via Baileys · Email via nodemailer + IMAP · SMS via Twilio · WebChat via WebSocket · Webhooks via HTTP · LINE via Messaging API · Reddit via OAuth2 · Matrix via Client-Server API

Smart Contracts & Multi-Language

Any Language. Plus Purp SCL.

PAW agents execute JavaScript, TypeScript, Python, Rust, and any language you throw at them via the api, js, and system execution modes. But PAW is the only agent framework that also ships with a Solana-native smart contract language built in — Purp SCL v1.1. Write, compile, and deploy Solana programs without leaving the chat. 

program TokenVault {
}

account VaultState {
  owner: pubkey
  balance: u64
  is_locked: bool
}

instruction Deposit {
  accounts:
    #[mut] vault_state
    #[signer] depositor
  args:
    amount: u64
  body:
    require(amount > 0, InsufficientFunds)
    vault_state.balance += amount
    emit(DepositMade, {
      depositor: depositor.key,
      amount: amount
    })
}

event DepositMade {
  depositor: pubkey
  amount: u64
}

error VaultErrors {
  InsufficientFunds = "Not enough funds"
  Unauthorized = "Only the owner"
}

Compiler Pipeline

.purp → Parse → Validate → Anchor Rust + TypeScript SDK + IDL
  • ✓ Full type system: u8–u128, i8–i128, f32, f64, bool, string, pubkey, bytes
  • ✓ v1.1 operators: ** exponentiation, ?? nullish coalescing, ... spread
  • ✓ Account attributes: #[mut], #[signer], #[init] with auto-generated context structs
  • ✓ SPL token operations: transfer, mint_to, burn, approve, revoke
  • ✓ Anchor-compatible Rust output with proper CPI patterns
  • ✓ TypeScript SDK with typed methods per instruction
  • ✓ Anchor-compatible IDL JSON generation
  • ✓ Purp.toml project config with dependency management
  • ✓ Backward compatible with legacy JSON format and v0.3 syntax
View Purp SCL ↗
Differentiators

Why PAW?

Safety-first architecture where the LLM plans and the system executes.
No exceptions, no shortcuts, every action traceable.

🛡️ Safety is Non-Negotiable

The validation pipeline runs on every action, in every mode. You can make it less intrusive with autonomous mode, but you can't turn it off.

🧱 Strict LLM/Execution Split

The LLM produces a JSON plan. The system validates it, then executes. The LLM never touches tools directly — eliminating an entire class of prompt injection attacks.

🤖 Risk-Aware Autonomy

Autonomous doesn't mean unsafe. Low and medium risk auto-execute. High risk is configurable. Critical actions always require confirmation.

📝 Full Audit Trail

Clawtrace logs every phase of every action with automatic secret redaction. Reconstruct exactly what happened, what the LLM reasoned, and why.

🔧 Intelligent Self-Healing

When an action fails, PAW diagnoses the failure type, determines if it's recoverable, applies a fix strategy, and only escalates when it genuinely can't recover.

⛓️ Purp SCL First-Class

Parse .purp files, validate, compile to Anchor Rust, generate TypeScript SDKs and IDL — all within the agent pipeline. No SCL integration like it.

🧠 Semantic Memory

Vector memory with TF-IDF scoring, scoped by session, user, or global, persisted to disk. Agents remember what matters across conversations.

🔌 Extensible by Design

Skill files, MCP protocol, custom tools, DAG workflows, multi-agent orchestration. PAW grows with your needs without code changes.

Native Dashboard

Real-Time Agent Control

A built-in web dashboard served by the Gateway — monitor agents, chat in real-time, switch modes, and inspect every action as it happens.

🐾 PAW Dashboard v3.0
Connected 🌙
Stats
Actions
247
Uptime
99.8%
Mode
01 Supervised
02 Autonomous
03 Locked
Pipeline
① Parse → ② Risk → ③ Validate
④ Approve → ⑤ Execute → ⑥ Audit
Agent Chat autonomous
You
Scrape the latest Solana TVL from DefiLlama
Agent
Fetching DefiLlama API for Solana chain TVL data…
✓ Action validated · risk: low · auto-approved
Agent
Solana TVL: $8.42B — up 3.7% in the last 24h. Data saved to memory.
You
Post that to Discord
Agent
Message sent to #sol-updates. Awaiting delivery confirmation.
Type a message…
Send
Activity Log Clear
14:32 ✓ discord.send completed
14:32 → risk:low auto-approved
14:31 ⚙ web.scrape DefiLlama…
14:31 ✓ web.scrape completed
14:30 → pipeline: 6/6 passed
14:28 ⚠ rate-limit approaching
14:27 ⚙ memory.store saving…
14:27 ✓ memory.store completed
💬

Live Chat Interface

Talk to your agents over WebSocket. Messages appear instantly with user/agent/system bubbles.

🔀

Mode Switching

Toggle between supervised, autonomous, and locked modes on the fly — no restart needed.

📊

Real-Time Stats

Action count, uptime percentage, and pipeline status update live via WebSocket push.

📜

Activity Log

Every action logged with timestamps and color-coded severity — info, success, warning, error.

🎨

Dark & Light Themes

Full theme toggle with persistent preference. Purple brand palette across both modes.

🔌

Zero Config

Served automatically by the PAW Gateway. Open your browser and the dashboard is ready.

Get Started

Up and Running in Minutes

Clone, configure, run. One AI API key and one channel token to start. 

# Clone the repository
$ git clone https://github.com/DosukaSOL/paw-agents.git
$ cd paw-agents

# Install dependencies
$ npm install

# Configure your API keys and channel tokens
$ cp .env.example .env

# Build and start
$ npm run build
$ npm start

Requires Node.js 20+, at least one AI API key (OpenAI or Anthropic), and at least one channel token.

AI Providers

Bring Your Own Model

Six providers with automatic failover. Configure your preferred model — use one or all.

OpenAI
GPT-4o · GPT-4 Turbo
Anthropic
Claude Sonnet 4
Google AI
Gemma 3 27B
Mistral
Mistral Large
DeepSeek
DeepSeek Chat · R1
Groq
Llama 3.3 70B

Automatic failover between all providers · ModelProvider interface for extensibility