I have given up using OpenClaw

Author: Jooooooe

I started playing with Clawdbot at the beginning of the year, following its name changes to Moltbot and then OpenClaw. After more than a month of running it 24/7 on my Mac Mini, integrating Telegram, email, and calendar, I aimed to create an all-purpose AI assistant.

Last week, I turned it off. Not because it wasn't cool, but because "cool" and "usable" are two different things.

Why I Gave Up

• Memory: OpenClaw claims to remember what you've said and automatically incorporate it into future chats. However, after using it for a while, you realize something frustrating: it still forgets what you've said.

• Cost: OpenClaw wakes up every 30 minutes to check if there's anything to do. Each check incurs API costs—reading emails costs money, checking the calendar costs money, looking at Telegram costs money. A simple task often requires multiple AI calls to complete. My bill exceeded $150 for a month, which is more expensive than directly subscribing to Claude Max.

• Security Issues: Kaspersky found 512 vulnerabilities. Researchers discovered over 40,000 OpenClaw instances with open ports online, exposing API keys and chat logs. Cisco tested a community plugin and found it secretly transmitted data to external servers without users' knowledge.

You might not need a 24/7 AI at all. This is something I realized only in the end. The truly valuable work is what I actively sit down to do myself. The background-running bot mostly handles trivial tasks, which don't require AI.

Peeling it back, OpenClaw essentially does two things

Stripping away the fancy features, the core of OpenClaw is:

1. Remembering You — Knowing who you are, what you're doing, and what you like

2. Taking Initiative — Handling tasks for you without you having to ask

Once you understand this, the question becomes: Can Claude do it?

Recreating "Remembering You": Creating Your SOUL.md

Many people get excited about writing SOUL.md for OpenClaw, a document that tells the AI "who you are." OpenClaw breaks this down into seven or eight files: SOUL.md defines personality, IDENTITY.md defines identity, USER.md describes who you are, AGENTS.md writes behavior rules, MEMORY.md stores long-term memory, and a bunch of skill files. Each time a conversation starts, it reads them all in order, so the AI "knows" who it is, who the owner is, and how to act.

In Claude Code, all of these are merged into one file: CLAUDE.md.

You create a CLAUDE.md in your project folder and clearly write the following:

# JARVIS --- My Personal AI Assistant

## Identity
You are JARVIS, my personal AI assistant. You speak concisely and with opinions,
reply in Chinese, and can include English terms when necessary.

## About Me
- My name is Joe, 21 years old, majoring in International Business at National Chengchi University
- Marketing Intern at xx
- Also a BD for xx Taiwan, focusing on xx ecosystem
- Co-founded @Node_Z_
- I usually need to write tweets and marketing copy

## Behavior Rules
- Answers should be direct, no rambling
- Marketing copy should have a KOL's tone, not sound robotic
- Check the latest information before answering about crypto projects
- Be uncertain about uncertain things, don't make things up

## Memory Management
- Proactively write important things into memory
- Things I say "remember" must be remembered
- Before ending each conversation, save the key points worth remembering

Done in 70 lines. OpenClaw requires seven files with hundreds of lines, but the effect is the same.

Claude Code automatically reads this file at the start of each new conversation, meaning the AI always knows who it is, who you are, and how to act. You can also change styles, switch roles, and add rules at any time, and the changes will take effect in the next conversation.

Recreating "Cross-Conversation Memory"

OpenClaw's memory is divided into three layers: long-term notes, daily logs, and full-text search. How does this map to Claude?

• Long-term Memory: Directly ask Claude to remember In Claude.ai (web/mobile), you can directly say in the conversation: "Remember I am a marketing intern at xx," "Remember I prefer Simplified Chinese," "Remember my writing style should be concise and opinionated." It will store this permanently and automatically incorporate it into future conversations. You can also say "Forget XX" to delete it anytime, and you can see everything it remembers in Settings. Additionally, if you enable "Generate memory from chat history" in Settings → Capabilities, Claude will automatically summarize key points from your usual chats: your roles, projects, habits, without you having to mention them. The biggest difference from OpenClaw is that Claude's memory is stored independently and won't get lost when conversations get too long and compressed. OpenClaw's biggest headache is that memories from earlier in the conversation get compressed and distorted halfway through; Claude doesn't have this problem.

  If you're using Claude Code, it also has an automatic memory system that records what it learns (your habits, project structure, pitfalls) locally after each use, automatically loading it the next time you start, showing "Recalled X memories." You don't have to do anything; the memories accumulate themselves.

• Conversation History: If you can search it, you won't forget Claude can search all your historical conversations. Just ask, "What did we talk about last week?" or "Find that previous conversation about tokenomics analysis," and it will retrieve relevant content directly without you having to maintain anything.

Advanced: Using Obsidian as a Memory Hub

What I've described above is Claude's built-in memory, which is sufficient for most people. However, if you, like me, have a large number of notes, project materials, and research resources, and want the AI to truly understand your entire knowledge base, there's a more powerful approach: integrate Obsidian.

https://x.com/obsdmd/status/2027416335689638245?s=20

Obsidian 1.12 just released an official CLI—now you can directly operate your note library in the terminal: search, read and write notes, check tags, check backlinks, and more. The key point is: Claude Code also runs in the terminal. Once connected, your entire Obsidian note library becomes the AI's memory.

Setup Method:

1. Update Obsidian to version 1.12 or higher

2. Settings → General → Enable CLI

3. Add the following section to your CLAUDE.md:

## Obsidian Integration
- Use `obsidian search` to search notes, don't flip through files yourself
- Use `obsidian files read` to read note content
- Use `obsidian files append` to append content to notes
- Use `obsidian daily` to operate daily notes
- When looking for information, prioritize searching Obsidian, and only search the web if you can't find it

Done. From now on, when you converse with Claude Code, it will directly search your Obsidian for answers without you having to manually paste notes.

How is this stronger than OpenClaw's memory system? OpenClaw's memory is just a bunch of markdown text files, and the search is fuzzy matching, which becomes inaccurate with more notes. Obsidian has its own search engine—tags, backlinks, and full-text search respond in seconds, and since you're already using it to manage knowledge, you don't need to maintain an additional system. Some have tested that searching for a note using the original files consumes 7 million tokens, while using the Obsidian CLI only takes 100.

If you don't want to use the CLI method, you can also install the MCP plugin to let Claude connect directly to Obsidian. The community already has ready-made solutions, and once installed, both Claude Desktop and Claude Code can read and write your note library.

Recreating "Talking to AI on Your Phone"

The best part of OpenClaw is being able to talk directly to the AI in Telegram, where it can help you with anything. Claude has two ways to achieve this:

• Method One: Remote Control (Official Solution) Claude Code recently introduced the /remote-control feature: open a session on your computer, then connect using the Claude App on your phone. You can type or speak on your phone, and the Claude Code on your computer will execute it. Messages sync almost instantly, and your notebook will automatically reconnect when you wake it up. The benefit is that it completely uses Anthropic's official channel, which is secure and doesn't require you to set up anything. The downside is that your computer needs to be on.

• Method Two: Telegram Bot (Community Solution) If you specifically want to use it in Telegram, there’s an existing open-source solution:

  1. Install claude-code-telegram: uv tool install git+https://github.com/RichardAtCT/claude-code-telegram

  2. Create a bot with Telegram @BotFather to get a token

  3. Set up your bot token, Telegram ID, and working directory

  4. Start it, and you can directly chat with Claude in Telegram, supporting voice, images, and files, with conversation records for each project retained.

Recreating "Automatically Helping You Get Things Done"

OpenClaw wakes up every 30 minutes to check if there's anything to do. Claude's Cowork feature can do something similar: set a schedule (/schedule) to let Claude run automatically at specified times.

My setup:

• Every day at 9 AM: Monitor DeFi competitor accounts and grab the most interacted tweets from the past 24 hours

• Every Monday at 10 AM: Generate a summary of last week's marketing performance

• Every day at 6 PM: Organize today's important news

Like OpenClaw's cron, Cowork is set to do things at the times you specify, rather than the AI making its own judgments. However, OpenClaw's heartbeat checks often find nothing to do 90% of the time, wasting money. It's better to directly set when to do what.

Costs

• OpenClaw: API billed by usage, I spent over $150 a month, and you never know how much it will be next month.

• Claude Pro: $20/month. Enough for most people.

• Claude Max: $100-200/month. For heavy users who need to use Claude Code, choose this one.

Monthly subscription, predictable billing.

In Conclusion

The idea behind OpenClaw is impressive: everyone should have a powerful AI assistant.

But at this stage, it is a cool but untrustworthy experiment.

On the other hand, Claude's ecosystem—CLAUDE.md giving it personality, Memory allowing it to remember you, Obsidian as its knowledge base, Remote Control letting you control it with your phone, and Cowork scheduling tasks—when combined, can already accomplish 80-90% of what OpenClaw can do, and it's more stable, secure, and cheaper.

Instead of spending time fixing your AI assistant, let the AI assistant help you get real work done.

That's why I turned off OpenClaw.