🐾 Local-First AI Second Brain

One Brain. Every Model.
Every Channel.

ClawPad gives you a serious desktop AI workspace that starts in local mode instantly, then scales to cloud only when needed. No account required. No forced subscription. Deterministic local + cache + cloud routing keeps quality high and costs predictable.

Download v0.7.5 Free Try in Browser
🔒 Local-first by default — chat immediately with on-device AI and user-owned memory
🔄 Deterministic model control — global default + per-topic override + runtime validation
🧠 Prompt Intelligence + Agent Jobs — command shortcuts, history recall, proactive execution
🧩 47 curated MCP extensions + Remote API for automation, integrations, and workflows

See ClawPad in Action

Latest 3 uploads plus top 2 most-viewed videos

Latest

How I Cut AI Costs with OpenClaw + ClawPad

Latest 20-slide walkthrough: free local intelligence, OpenClaw setup, memory, channels, MCP servers, and cloud-only-for-hard-task routing.

⏱️ New Upload

ClawPad + OpenClaw Setup in 4 Minutes (Gmail + Calendar MCP Live Demo)

Step-by-step quick setup showing OpenClaw integration with Gmail and Calendar MCP in a real working flow.

⏱️ New Upload
Latest

OpenClaw + Advanced Local Model Benefits — Why I Stopped Paying for AI Tools

Advanced local-model deep dive: private local workflows, stronger memory behavior, and practical cloud escalation only when needed.

⏱️ New Upload
Most Viewed

OpenClaw vs $200/mo AI Tools: Token Efficiency + Huge Savings

Most-viewed upload on the channel: why local-first routing with OpenClaw + ClawPad can cut recurring AI costs significantly.

⏱️ Popular
Most Viewed

AI Second Brain — Full Tour

Second most-viewed upload: full product tour covering OpenClaw embedding, setup flow, and daily usage across model providers.

⏱️ Popular
Download ClawPad v0.7.5 Free Try Free in Browser More Videos on YouTube

No download needed — try local AI directly in your browser

Why ClawPad?

ClawPad is your AI workspace and gateway. Instead of being trapped inside one vendor's chat app, you get a secure local brain that works across models and channels, retrieves the right context per project, and extends via an ecosystem of tools. The result: higher output, lower friction, and full control over your data.

🆓

Completely Free — For Everyone

Free for personal, business, and commercial use. Remote API usage is included and free. No subscription fees, no hidden costs, no premium tiers. You only pay providers directly when you choose cloud models.

🔒

Local-First, User-Owned Memory

Your long-term context lives on your device, encrypted and portable. Memory is scoped per workspace — not a leaky global soup. Full control: view, edit, delete, export your memories anytime. Your AI, your data, your control.

🔄

Multi-Model by Design

Switch between OpenAI, Anthropic, Google, and Grok seamlessly. Use the best model for the job: reasoning, coding, speed, cost. Standardized tool + memory layer means your workflow doesn't break when you change models. Model-agnostic. Workflow-consistent.

🐾

One Brain Across Channels

Bring the same ClawPad brain into Telegram, Discord, Slack, LINE, and WhatsApp. Conversations are not isolated islands — ClawPad keeps continuity across devices and channels. Your assistant follows you everywhere.

🔐

Your Data, Your Control

Unlike cloud-based solutions, all your conversations stay on YOUR device. API keys encrypted at rest with machine-specific keys — no cloud, no keychain dependencies.

📂

Project-Scoped Context Injection

ClawPad doesn't just stuff the prompt. It retrieves the right context: repo files, docs, decisions, APIs, previous sessions. Workspace isolation prevents cross-project contamination. Always relevant. Never noisy.

🧠

AI That Actually Remembers

Unlike web-based chat interfaces, ClawPad's memory system means your AI remembers your projects, preferences, and past conversations across sessions. Memory survives compaction and works across all channels.

🔌

Bring Your Own API Key

Use your own Anthropic, OpenAI, Google, or Grok API key. Four providers, 27+ models. No middleman. Direct connection means better privacy and you see exactly what you pay for.

💻

Native Desktop App

Not a slow web wrapper. ClawPad is a proper native application that's fast, responsive, and integrates with your system's features.

📂

Organize Everything

Create topics for different projects. Attach files, images, PDFs. Keep your AI conversations as organized as your file system.

🎨

10 Beautiful Themes

10 stunning themes inspired by great artists: Aurora Teal, Starry Night, Klimt Golden, Monet Lilies, Paper Beige, and more. Switch instantly with no restart.

🔍

Accessibility First

Scale the entire UI from 75% to 200% with Cmd/Ctrl +/-. Every element scales proportionally for comfortable viewing on any screen.

🔎

Cross-Project Intelligence

ClawPad's FTS5 knowledge engine indexes every conversation. When you ask a question, it automatically finds relevant context from ALL your topics — past projects inform current work.

🎛️

Configurable AI Pipeline

Fine-tune how much context the AI uses with intuitive sliders. Control token budgets for history, knowledge engine, and memory. Balance intelligence vs. cost — everything is on by default.

⚡

4-Layer Token Efficiency Architecture

ClawPad's architecture uses four core optimization layers: intent classification, context elision, answer-directed query distillation, and progressive conversation summarization. Supporting modules like semantic cache and prompt caching further reduce cloud spend.

🧠

Semantic Cache with MiniLM

Local MiniLM embeddings detect semantically identical questions — "fix login bug" matches "repair auth issue" at 85% similarity. 3-5x better cache hit rates than plain text matching. Zero API cost for repeated queries.

🎯

Smart Per-Topic Model Selection

Each topic remembers its own model. Use GPT for code generation, Claude for architecture, a cheaper model for quick questions. Switch with one click — context preserved, no configuration needed.

🐾 OpenClaw Embedded

What is OpenClaw? OpenClaw is a local AI gateway that runs on your computer. It provides unified memory across all your AI interactions, whether from ClawPad desktop, Telegram, Discord, Slack, LINE, WhatsApp, or any connected channel. Think of it as your personal AI brain that never forgets.

  • ✓ Automatic installation — no manual setup required
  • ✓ Runs locally on your machine (port 18789)
  • ✓ Syncs context between ClawPad and all connected channels
  • ✓ Encrypted storage for API keys and data
  • ✓ Zero cloud dependency for your private data

How It Works

OpenClaw is your personal AI gateway that keeps everything in sync

1
🖥️

Chat on ClawPad

Start a conversation on your desktop. Organize by topics, attach files, get work done.

→
2
🧠

OpenClaw Remembers

Your AI gateway stores context, memories, and conversation history locally.

→
3
📱

Continue on Any Channel

Pick up where you left off on Telegram, Discord, Slack, LINE, or WhatsApp. Your AI knows what you discussed.

🖥️ ClawPad (Desktop)
Help me plan the product launch for next week
I'd be happy to help plan your launch! Let's break it down...
Focus on the marketing timeline
Here's a day-by-day marketing timeline for your launch...
🔄
Synced via OpenClaw
📱 Telegram / Discord / WhatsApp
What about social media posts?
Building on our launch timeline, here are social media posts for each day...
Perfect! Add Instagram stories too
Great idea! I've added Instagram story suggestions to complement...

✨ Same conversation, different devices. Your AI remembers everything because OpenClaw keeps your context unified.

Your AI on Every Messenger

Access your AI assistant wherever you chat — same memory, any platform

Telegram

2B+ users worldwide

Available Now

WhatsApp

2B+ users worldwide

Available Now

Messenger

1B+ users worldwide

Coming Soon

LINE

200M+ users in Asia

Available Now

WeChat

1.3B+ users in China

Coming Soon

Discord

150M+ active users

Available Now

Slack

30M+ daily users

Coming Soon

Signal

40M+ privacy users

Coming Soon
🚀

One AI Brain, Every Platform

No matter which messenger you use, OpenClaw keeps your AI context unified. Start a conversation on your desktop, continue on Telegram during your commute, switch to Discord at home, or pick up on LINE — your AI remembers everything.

One Interface, Multiple AI Models

Switch between the world's best AI models with one click

Claude

Anthropic's most capable AI

Recommended

GPT-5.2

OpenAI's powerful model

Available

Gemini

Google's multimodal AI

Available

Grok

xAI's conversational model

Available

Per-topic model selection: Each topic remembers its model. Use GPT-5.2 Codex for code generation, Claude Opus for architecture decisions, a cheaper model for quick questions. Switch with one click — context preserved across all channels.

Token Efficiency Engine

4 architecture-backed optimization layers, plus semantic cache and prompt caching for lower cloud-token usage.

90%
Prompt Cache Savings

Anthropic cache control reuses system prompts across turns

3-5x
Better Cache Hit Rates

MiniLM semantic embeddings catch similar questions

Policy
Local-First Routing

Simple tasks stay local; cloud is used when complexity or tool needs exceed local limits

4
Core Architecture Layers

Intent classification, context elision, query distillation, and progressive summarization

🎯

Layer 1: Intent Classification

Each message is classified as LIGHT/MEDIUM/HEAVY so context depth and budgets match task complexity.

🧾

Layer 2: Context Elision

Follow-up turns avoid re-injecting unchanged context, reducing repetitive token spend.

🧭

Layer 3: Query Distillation

Local assist creates answer-directed directives that steer cloud responses toward concise, task-shaped outputs.

🗜️

Layer 4: Progressive Summarization

Long conversations are compacted into rolling summaries to preserve continuity without runaway token growth.

🧠

Semantic Cache

MiniLM embeddings detect similar questions and reuse prior answers with zero API cost on cache hits.

🔄

Prompt Caching

Provider prompt-caching features reduce repeated system/context token charges on long threads.

📏

Exact Token Budgeting

Per-model context/output limits and token counting keep payloads within safe, efficient bounds.

🔀

Differential Context Injection

Context headers and history are injected only when needed, reducing repeated prompt overhead.

Local Model Engine

Default local runtime (`local_openai` + qwen3-4b-q4), gate-scored local/cloud routing, and semantic cache orchestration built directly into ClawPad.

🧠

Embedded Inference Runtime

Local models run in-process via llama-cpp-python (no external server dependency). This reduces moving parts and improves stability for daily use.

🧭

Intent + Context Aware Router

Each query is classified (LIGHT/MEDIUM/HEAVY), scored for context-fit/tool-need confidence, then routed to local or cloud. Topic-level model overrides always win over global defaults.

⚖️

Local First, Cloud When Needed

Simple requests use local models for speed and privacy. Hard or tool-intensive requests auto-escalate to cloud models for quality and capability continuity.

Lower Cost

Local-first handling + semantic cache reduce paid API calls while preserving cloud quality for complex work.

Lower Latency

In-process local inference avoids network round trips for lightweight tasks and routine follow-up questions.

Better Privacy

More requests can stay on-device when routed locally, keeping sensitive project context under user control.

Higher Reliability

If a request exceeds local capability, ClawPad escalates to cloud and preserves model-interface consistency instead of producing unstable output.

Supported Local Models (Q4 GGUF)

GLM-4.7 Flash

Min RAM: 24GB • Effective Context: 6,144 • Size: ~18.3GB

GPT-OSS 20B

Min RAM: 16GB • Effective Context: 6,144 • Size: ~11.6GB

Qwen3 14B

Min RAM: 16GB • Effective Context: 6,144 • Size: ~8.4GB

Qwen3 8B

Min RAM: 10GB • Effective Context: 6,144 • Size: ~4.7GB

Qwen2.5 7B

Min RAM: 10GB • Effective Context: 6,144 • Size: ~4.4GB

Qwen2.5 Coder 7B

Min RAM: 10GB • Effective Context: 6,144 • Size: ~4.4GB

Mistral 7B v0.3

Min RAM: 10GB • Effective Context: 6,144 • Size: ~4.1GB

Qwen3 4B

Min RAM: 6GB • Effective Context: 3,072 • Size: ~2.5GB

Phi-3.5 Mini

Min RAM: 6GB • Effective Context: 3,072 • Size: ~2.2GB

Llama 3.2 3B

Min RAM: 6GB • Effective Context: 6,144 • Size: ~2.0GB

Gemma 3 4B

Min RAM: 6GB • Effective Context: 6,144 • Size: ~2.5GB

Qwen2.5 1.5B

Min RAM: 4GB • Effective Context: 3,072 • Size: ~0.9GB

SmolLM2 1.7B

Min RAM: 4GB • Effective Context: 1,536 • Size: ~1.0GB

All local models run with compact-context optimization and deterministic handoff: local-first, semantic cache second, cloud path when complexity/tool requirements exceed local limits. Tool-capable local models include GLM-4.7 Flash, GPT-OSS 20B, Qwen3 14B, and Qwen3 8B; reliable MCP local execution is currently validated on Qwen3 14B.

🌐

Try Local AI Right Now — In Your Browser

Run Phi-3.5, Qwen2.5, Llama 3.2, or Gemma 2 directly in your browser via WebGPU. No download, no API key, completely free.

Launch Browser Demo

Agent Jobs: Major Workflow Engine

Agent Jobs is ClawPad's always-on execution layer: scheduled work, manual run-now dispatch, run history, structured outputs, and health-aware recovery built into the product core.

⏱️

Dual-Timer Runtime

Two execution loops run together: full proactive analysis on a 5-minute cadence and a lightweight 30-second heartbeat pulse for fast due-job detection and health transitions.

🎯

Exact Dispatch + Run Now

Due-job execution honors explicit job_id to prevent wrong-job dispatch, and Run Now triggers real manual engine execution (not detail-only navigation).

📈

Run-Scoped History

Job definitions are separated from run instances. Each run tracks status, trigger source, timeline updates, and canonical structured_report payloads for dashboard/report rendering.

🗂️

Lifecycle Controls

Manage jobs through active, archived, and trash states with restore/clone/delete flows, plus run-aware timelines and analytics surfaces in dashboard views.

🩺

Health-Aware Scheduling

Required MCP dependencies are checked before execution. Jobs can skip safely when dependencies are unhealthy, retry on later pulses, and avoid silent degradation.

🪝

Automation Hooks

HTTP hooks support wake, agent turn, proactive cycle, pulse, manual job run, and status audit, enabling CI/cron/monitoring systems to trigger and observe Agent Jobs externally.

Remote control: drive Agent Jobs via /api/proactive/* and /api/hooks/job/run, monitor with /api/hooks/status and /health, and keep desktop + channel execution context in sync.

Release-Grade Architecture

Mapped to `REQUIREMENTS.md` + `ARCHITECTURE.md`, then validated by ClawPad's RRG go/no-go gate before shipping.

🧬

Model Interface Contract

Global setting inheritance with per-topic overrides and runtime model validation avoids broken provider/model combinations in channel flows.

🧠

Local-First by Default

`local_openai` + qwen3-4b-q4 runs out of the box. Cloud models are optional upgrades for heavy reasoning and tool-intensive tasks.

🛰️

Shared Brain Across Channels

Desktop + Telegram + Discord + Slack + LINE + WhatsApp use one shared memory and channel identity mapping, so context remains coherent across surfaces.

🧪

RRG Release Gate

Fast/API/full tier matrices, release regression suite, and artifact preflight checks run as the official ship/no-ship decision path.

🔐

Signed + Notarized Distribution

Every release DMG is Developer ID signed, notarized, stapled, checksumed, and aligned with updater asset naming conventions.

📦

Update Path Integrity

Website download links, GitHub release assets, and updater resolution are preflight-validated to prevent broken update prompts.

Powerful Features

25 powerful features for serious AI work — 100% Free

⚡

4-Layer Token Efficiency Engine

Intent classification, context elision, query distillation, and progressive summarization are the core architecture layers. Semantic cache and prompt caching further lower cloud-token usage.

🎯

Hybrid Local + Cache + Cloud Routing

ClawPad routes each query through a token-efficient pipeline: local model first for fast/private tasks, semantic cache layer for repeat or similar questions, then cloud model for heavy/tool-intensive work. Per-topic model overrides are still respected.

🧩

47 MCP Extensions

One-click install MCP servers from the MCP Server Store. Filesystem, Git, Memory, Web Fetch, YouTube, Playwright, SQLite, Notion, Slack, Home Assistant, and more. Every server gives your AI new real-world tools.

🧠

Smart Context Injection

ClawPad's superpower. Every message includes topic name, recent history, and relevant files — the AI always knows exactly what you're working on.

🔄

Multi-Model Support

Switch between Claude, GPT, Gemini, and Grok in one click. Global defaults propagate across channels, while per-topic model overrides remain isolated and enforced.

📁

Topic-Based Organization

Each topic is an isolated workspace with its own history, files, and context. Work on multiple projects simultaneously without mixing context.

📋

Toggle Copy & Multi-Copy

Click any message to copy it. Click again to deselect. Select multiple messages and they concatenate on your clipboard. Code blocks get their own copy button too.

⚡

Parallel Conversations

Start a chat, switch topics, and both continue streaming. Work on multiple things at once with visual indicators for active responses.

📎

File Management

Attach files to topics. Track inputs and AI-generated outputs with version history. Drag & drop to add, click header to browse.

🔍

Cross-Project Knowledge Engine

FTS5 full-text search indexes every conversation. The AI finds relevant context from ALL topics — your past projects automatically inform current work.

💾

Long-Term AI Memory

Your AI remembers facts, preferences, and decisions across sessions. Survives conversation compaction, works across all channels. Auto-enabled with zero setup.

🎛️

Supported Local Models

GLM-4.7 Flash, GPT-OSS 20B, Qwen3 14B/8B/4B, Qwen2.5 7B, Qwen2.5 Coder 7B, Qwen2.5 1.5B, Mistral 7B v0.3, Phi-3.5 Mini, Llama 3.2 3B, Gemma 3 4B, and SmolLM2 1.7B.

🔗

Unified Memory

Seamless sync across all channels via OpenClaw Gateway. Start on desktop, continue on any connected channel. Your AI remembers everything.

🔒

Private & Secure

Local SQLite database, encrypted API keys, no cloud storage. Your data stays on your device. Works offline for browsing history.

🎨

10 Art-Inspired Themes

Aurora Teal, Starry Night, Klimt Golden, Frozen Steel, Monet Lilies, Paper Beige, and more. Real-time switching with desktop-grade color clarity.

💻

Native Performance

Built with PyQt6 — not Electron. Fast startup, smooth scrolling, proper macOS integration. Feels like a real app because it is one.

🔧

Full Control

Database backup/restore, connection management, API key configuration, device info, and automatic updates. You're in charge.

✨

Code, Markdown & Tables

Full syntax highlighting, markdown rendering with native table support, per-block copy buttons, and toggle copy on messages. Markdown tables render with themed colors and alignment.

🖥️

Smart HTML Rendering

AI responses render as rich interactive HTML — dashboards, styled data tables, color-coded charts, and SVG graphics right inside chat bubbles. Not screenshots — real, live rendering.

♾️

Auto-Continuation

Never truncated again. When responses hit the token limit, ClawPad auto-continues in a new bubble. Up to 5 continuations for 80K+ token responses — zero manual intervention.

🤖

Agentic AI Tools

OpenClaw's agentic loop gives your AI real tools — file search, code reading, structured analysis. Every tool call shows as a visible card. Visual control with permission presets and per-tool allow/ask/block.

🌍

12-Language Interface

Full internationalization with 12 languages: English, Korean, Japanese, Chinese (Simplified), Chinese (Traditional), Spanish, French, German, Portuguese, Italian, Russian, and Arabic. Switch instantly in Settings.

🛠️

Claude Code Tool Parity

Full parity with Claude Code's tool suite — Edit, MultiEdit, LS, NotebookEdit, TodoWrite, and TodoRead. Your AI agent has the same capabilities as a professional coding assistant.

⌨️

Keyboard Shortcuts

Comprehensive keyboard shortcuts in the menu bar for power users. Navigate topics, send messages, toggle panels, and manage conversations without touching the mouse.

⬆️

Input History Recall

Terminal-style input history with Up/Down arrow keys. Recall and re-send previous messages instantly — just like in a Unix shell or IDE terminal.

Why Download ClawPad Today

Everything you need to start now: local-first AI, serious desktop workflow controls, and zero lock-in licensing.

🚀

Start in Under a Minute

Install, open, and chat immediately in local mode. No API key, no account setup, no cloud dependency required.

🧠

Prompt Intelligence Built In

Use shell-style shortcuts, live similar-prompt suggestions, and cross-channel prompt history search from one input system.

📌

Agent Jobs for Repeat Work

Run scheduled and manual AI jobs with lifecycle control, progress visibility, health checks, and channel-connected execution.

🔌

Developer-Grade Automation

REST + WebSocket APIs, webhooks, and job endpoints let you integrate ClawPad into your own internal and external systems.

Developer Remote API

Build on top of ClawPad — 180 endpoints across 15 categories for conversations, files, agent jobs, UI automation, MCP servers, and gateway control.

Full API reference is available at /api/.

🔌

REST API + WebSocket

Full HTTP API on port 18790 with real-time WebSocket events. The API auto-starts by default (remote_api_autostart=true), and can also run in explicit API/headless modes for automation.

📡

Webhook HTTP Ingress

Trigger wake/agent/proactive flows via /api/hooks/* and audit delivery in webhook status logs for production automation.

🧠

Agent Jobs APIs

Create/filter/select/schedule agent jobs through /api/proactive/* while keeping desktop + channel execution context aligned.

15 API Categories

💚 Health & Status 5
📚 Developer Docs & Discovery 12
📁 Projects 28
💬 Messages & AI 7
📎 Files & Artifacts 6
🔍 Search 4
⚙️ Settings & Config 16
🎨 Themes 2
🤖 AI Models 1
🔌 Gateway & Connection 46
🗄️ System 4
🕹️ UI Control 25
📸 Screenshots 8
🧩 MCP Servers & Tools 15
📡 WebSocket 1

Get Started in Seconds

bash Launch with API 
# Standard launch (API auto-starts by default)
./ClawPad.app/Contents/MacOS/ClawPad

# Headless automation runtime
./ClawPad.app/Contents/MacOS/ClawPad --stealth --api --api-port 18790

# Health check
curl http://127.0.0.1:18790/health
python Trigger Agent Job Cycle via Webhook 
import requests, uuid

BASE = "http://127.0.0.1:18790"
run_id = f"site-{uuid.uuid4().hex[:8]}"

# Ask ClawPad to run a proactive cycle now
hook = requests.post(f"{BASE}/api/hooks/proactive").json()
print(hook["success"], hook["data"]["run_id"])

# Read dashboard state
state = requests.get(f"{BASE}/api/proactive/dashboard/state").json()
print("cards:", len(state["data"]["cards"]))

# Inspect webhook execution log
status = requests.get(f"{BASE}/api/hooks/status").json()
print("recent hooks:", len(status["data"]["recent"]))

What Can You Build?

📊

Analytics Dashboards

Track AI usage, token consumption, and conversation patterns across all your projects.

🤖

Workflow Automation

Build CI/CD pipelines that use AI for code review, documentation, and testing.

🔗

Custom Integrations

Connect ClawPad to your existing tools — Notion, Jira, Slack, or your own apps.

🧪

AI Pipelines

Chain multiple AI calls with different models, contexts, and file inputs for complex tasks.

Get Started in 3 Minutes

ClawPad installs OpenClaw automatically — just download and go. Or try it free in your browser first →

1

Download ClawPad

Choose your platform and download ClawPad. Free for all use — personal, business, and commercial.

🍎 macOS — Available 🪟 Windows — Soon 🐧 Linux — Soon
2

Launch in Local Mode (No API Key Needed)

On first launch, ClawPad starts with its embedded local model by default. Add cloud API keys only when you want higher-end reasoning or provider-specific models.

Local default: qwen3-4b-q4 • Optional: sk-ant-... / sk-... / AIza... Optional cloud keys →
✨ OpenClaw Gateway + local runtime install automatically — no extra steps needed.
3

Connect Remote Access (Desktop + Mobile)

Use ClawPad on desktop and mobile PWA with one shared brain. Desktop stays passive by default and serves remote sessions only when you request a ticket.

a Telegram setup: In desktop ClawPad, link Telegram in Settings, then send /connect to your linked bot.
b Mobile remote access: In Telegram, send /ticket to your ClawPad bot.
c Open Mobile PWA: Tap the ticket link, or open /app/ and paste the ticket.
d Save to Home Screen: In PWA Settings, use Get the App to add the web app icon (PWA, not App Store build).
e Other channels: WhatsApp, Discord, Slack, and LINE all support in-app setup with the same shared memory.

That’s it. Desktop + mobile PWA + channels now share the same AI memory and topic context, including voice input and TTS read-aloud.

🎉

You're All Set!

Now you can:

  • 💻 Use ClawPad on your desktop for focused work
  • 📱 Use Telegram, Discord, Slack, LINE, or WhatsApp when you're on the go
  • 🧠 Enjoy shared memory — your AI remembers across all channels
  • 📁 Organize by topics — keep projects separate
  • 🔄 Seamlessly switch — continue any conversation anywhere

Troubleshooting

Having issues? Here are quick solutions

🔴 OpenClaw Gateway Won't Start

Try these steps:

  1. Open ClawPad Settings → OpenClaw → Restart Gateway
  2. Check if port 18789 is available
  3. On macOS: Allow in System Preferences → Security

🔑 API Key Not Working

Verify your key:

  1. Check at console.anthropic.com
  2. Make sure you have API credits
  3. Re-enter the key in Settings → API Keys

📱 Channel Not Connecting

Connection checklist:

  1. Make sure OpenClaw Gateway is running (green status)
  2. For Telegram: use the correct bot @OpenClawBot and try a new connection code
  3. For WhatsApp: re-scan the QR code from ClawPad's channel settings
  4. For Discord/Slack/LINE: check your bot token or credentials in channel settings

🔄 Conversations Not Syncing

Check sync status:

  1. Both devices must connect to the same OpenClaw
  2. Gateway must be running on your computer
  3. Check Settings → OpenClaw → Memory status

💾 How to Backup My Data

Your data location:

~/ClawPad/

Contains: database, projects, settings. API keys are encrypted with your machine's unique ID.

🔄 How to Reset Everything

Fresh start:

  1. Settings → OpenClaw → Reset OpenClaw
  2. Reset also clears local model/runtime/STT caches for a true clean reinstall
  3. Or delete ~/ClawPad/ folder
  4. Restart ClawPad to re-setup

🎨 How to Change Theme

Switch themes:

  1. Click Settings (gear icon) in sidebar
  2. Select from 10 themes (Starry Night is default)
  3. Theme applies instantly — no restart needed

🔍 How to Zoom In/Out

Scale the entire UI:

  1. Cmd/Ctrl + to zoom in (up to 200%)
  2. Cmd/Ctrl - to zoom out (down to 75%)
  3. Cmd/Ctrl 0 to reset to 100%

🎤 How to Use Voice Input + TTS

Speak, transcribe, and read aloud:

  1. Click the mic button to start local STT voice input (faster-whisper)
  2. Speak naturally, then stop recording to insert transcribed text
  3. Enable TTS in Settings → General → Text-to-Speech
  4. Click the speaker button on any finalized assistant message
  5. Use Ctrl+Shift+Space to toggle playback and Ctrl+Shift+S to toggle auto-read mode
  6. Highlighting follows playback smoothly with adaptive character-level sync
  7. On mobile PWA, use the same speaker controls and synchronized highlighting

Still stuck? For feedback, questions, and inquiries, post in the ClawPad YouTube Community.

Download ClawPad

Free for personal, business, and commercial use. Remote APIs are included at no extra cost. No account required.

💡 Not ready to download? Try ClawPad free in your browser — run local AI models with zero install.

macOS

v0.7.5 — macOS (Apple Silicon + Intel via Rosetta)

~352 MB

Windows

Coming Soon

Linux

Coming Soon

v0.7.5 Release Notes

Platform-focused release: stronger local-first routing, Prompt Intelligence, Agent Jobs maturity, and production-grade API workflows.

🧠

Prompt Intelligence

Input-level command shortcuts, live similar-prompt suggestions, and searchable prompt history now work as a single productivity layer.

⚙️

Deterministic Local/Cloud Routing

Routing and model inheritance were tightened so local-first behavior remains predictable while still escalating cleanly for complex tasks.

📌

Agent Jobs Runtime Improvements

Dashboard lifecycle behavior, scheduling reliability, and job state observability were upgraded for real daily operational workflows.

🧩

47-Extension MCP Catalog

The extension catalog experience now reflects broader official/community coverage for practical integrations across work and personal tasks.

🎨

Theme-Aware UI Consistency

Core workflows and views were normalized to respect theme colors and improve contrast/clarity across diverse visual themes.

↩️

Navigation + Input UX Hardening

Back navigation and prompt suggestion behavior were refined to reduce friction and make fast context switching reliable.

📡

Remote API + Webhook Readiness

API and webhook surfaces were further aligned with production automation use cases and release validation standards.

📦

Release Pipeline Reliability

Packaging, signing, and release checks were reinforced so downloadable builds remain stable and trustable for end users.

What's Included

🖥️

ClawPad Desktop App

Full-featured AI workspace with topics, files, themes

🧩

47 MCP Extensions

One-click install: files, git, browser, database, web fetch & more

🐾

OpenClaw Gateway

Local AI gateway that syncs everything

📱

Multi-Channel Integration

Telegram, Discord, Slack, LINE, WhatsApp — all set up inside ClawPad

🧠

Shared AI Memory

Your AI remembers across all platforms

🔊

Voice Input + TTS Read-Aloud

Local STT mic input, Edge TTS voices, auto-read mode, and adaptive character-level highlighting

🔍

Cross-Project Knowledge Engine

FTS5 search indexes all conversations across topics

💾

Long-Term AI Memory

Persistent facts and preferences across sessions

🎨

10 Beautiful Themes

Art-inspired themes including Aurora Teal, Starry Night, Klimt Golden, and Paper Beige

🔍

Zoom Scaling

Scale UI 75%-200% with Cmd/Ctrl +/-

🖥️

Smart HTML Rendering

Charts, tables, and rich content rendered natively in chat

♾️

Auto-Continuation

Unlimited AI response length with automatic continuation

📊

Markdown Table Rendering

Beautiful themed tables with colors and alignment

🤖

Agentic AI Tools

Real tools with visible tool cards and permission controls

🌍

12-Language Interface

Full i18n — English, Korean, Japanese, Chinese (Simplified/Traditional), Spanish & more

🛠️

Claude Code Tool Parity

Edit, MultiEdit, LS, NotebookEdit, TodoWrite & TodoRead