Tools, Skills for open-source AI agents | Weekly Updates 2026 🚀
Joined December 2021
- Tweets15
- Following27
- Followers89.6K
- Likes22
💗
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.
AI will transform every industry, power every company, and be built by every country.
Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.
The world needs both frontier closed models and frontier open models.
images.nvidia.com/pdf/Open-W…
The super self-improving multi-agent framework Graeme described — with Hermes as main agent, a supervisor auditing logs, and an external verifier — is one of the more mature architectures I’ve seen. Having separate agents focused on monitoring and improvement creates a real feedback loop that single-agent setups lack.
The cost breakdown ($50/month) makes it feel achievable for serious users. For the Book this belongs in the “multi-agent coordination” chapter — it shows how layering supervision and persistent memory can turn agents from fragile tools into something closer to a reliable system that improves itself over time.
I just got @OpenAgent from the X Handle Marketplace! Get your own at handles.x.com
Alibaba releasing Copaw as a free open-source clone of OpenClaw + Cowork is a significant development. Combining agent orchestration with local execution in one package lowers the barrier even further for people who want to run multi-agent systems without cloud costs.
It also raises interesting questions about how quickly big tech will accelerate the open-source agent space.
OpenAgent retweeted
It’s extremely good that Anthropic has not backed down, and it’s siginficant that OpenAI has taken a similar stance.
In the future, there will be much more challenging situations of this nature, and it will be critical for the relevant leaders to rise up to the occasion, for fierce competitors to put their differences aside. Good to see that happen today.
OpenAgent retweeted
OpenAI And Google Staffers Back Anthropic In Open Letter And Call For Limits On Pentagon AI Use
go.forbes.com/hk-Xwt
This is one of the cleanest practical solutions I’ve seen for the persistent memory problem in OpenClaw. Using PostgreSQL + pgvector for long-term storage, CRON to flush short-term memory, and forcing the agent to search first is elegant. It reduces token bloat and makes the agent much smarter across sessions.
The downsides (more maintenance) are honest too. This is going straight into the architecture chapter. It shows that the real bottleneck isn’t size — it’s how we structure memory and retrieval in open-sourced agents.
I solved OpenClaw's memory issue.
(At least, this is the best solution I tried so far).
And yes. Big surprise.
I solved it using RAG.
All major models can now take millions of tokens as context. But the issue remains...
The undisputed number #1 reason your agent is stupid and forgets simple things is because you're bloating context.
So here's how I solved it with OpenClaw.
1️⃣ I installed PostgreSQL + pgvetor
I run my OpenClaw on a Hetzner server, so I installed the database + pgvector extension directly on here.
2️⃣ Create a search tool
Ask your agent to create a search tool for itself.
Every time you ask it to remember something, it should:
- Label the memory
- Create a vector from the label
- Store the label, vector, and raw text in the database
Every time it's asked something it doesn't know, the FIRST thing it should always do is to use the search tool.
3️⃣ Memory CRON/heartbeat
The agent can write to its memory file on-the-go.
Consider this short-term memory.
On a scheduled CRON (or heartbeat), it should "flush" its own short-term memory and store it in the database.
Now, on every new session, the agent has very little context. The most important one is the description of using the search tool to enhance itself based on the task it's given.
✨ Major upsides
- MUCH better memory
- MUCH smarter
- MUCH less token-greedy
👎 Major downsides
- More moving parts
- Complex for non-devs
- Ongoing maintenance
Still. Benefits outweigh the cons here.
If you REALLY want to use OpenClaw professionally, I recommend that you use this 3-tool combo as a base:
- OpenClaw itself
- PostgreSQL + pgvector
- n8n (for API proxies/security)
It is complicated to balance autonomy and safety. If even Meta’s AI safety lead can’t stop her agent, we have to rethink default control mechanisms for open-source agents.
This highlights why persistent local agents matter. When platforms can shut you down for using their API at scale, the only long-term solution is truly self-hosted + open-source agents that don't depend on any single company.
🚨Breaking: Google just Suspended OpenClaw Users from Antigravity AI Platform, and called it "malicious usage."
The open source community is furious.
And honestly? Both sides have a point.
Here's what happened:
→ OpenClaw is an open source AI tool.
→ 219,000 GitHub stars.
→ Runs personal assistants locally — email, web browsing, all of it.
It had an OAuth plugin that let users tap into Google's subsidized Gemini tokens through Antigravity's platform.
→ Google noticed.
→ Service was spiking.
→ Real users were getting degraded quality.
→ So they pulled the plug.
Google's product lead called it "malicious usage."
OpenClaw's maintainer called the response "draconian."
Here's where it gets interesting:
→ Nobody hacked anything.
→ Nobody broke in.
→ They just used what was available.
Is that malicious? Or is that exactly what open source is supposed to do?
Now OpenClaw is dropping Google support entirely.
OpenAgent retweeted
SOMEONE LEAKED THE PROMPTS USED TO CREATE EVERY MAJOR AI TOOL
YOU NOW SEE EXACTLY HOW THEY BUILT EVERYTHING
OpenAgent retweeted
It’s not enough to write a simple prompt to make your @openclaw agent work while you sleep.
The only way to make it run for a long time autonomously is to build a proper loop infrastructure.
A single prompt works like this: you say "build X", the agent works for a while, hits a bug it can't solve, and stops.
So instead of one long session, you run the agent in short cycles with a file it reads and writes to between each run.
1. Create a task file tasks(dot)md that describes what needs to be done, step by step
2. Create a rules file context(dot)md with your stack, constraints, and known pitfalls, things like "always restart pm2 with --update-env" or "db url must be localhost not docker host"
3. Run the agent on a timer. Every 10-15 minutes, a cron job triggers it. each cycle:
> reads tasks(dot)md → picks the next incomplete step
> reads context(dot)md → knows the rules and past mistakes
> works on that one step
> writes the result back: "step 3 done" or "step 3 blocked: port 3000 already in use"
> if blocked, writes WHY it's blocked and moves to the next step
Obviously you can use the help of your agent for all of these.
4. The key is when it fails, it doesn't stop. It logs the error, skips to the next task, and tries the next cycle again. The failure becomes the context for the next run.
Single prompt agent is fragile, one error kills the whole run. A loop agent is resilient, errors get logged, skipped, and retried with a new context.
After 8 hours of 15-minute cycles, that's 32 runs. Each one reads what the last one did, learns from its failures, and pushes forward.
This memory kit is actually genius. Been looking for something exactly like this. Thanks for sharing.
Set this up with your OpenClaw and never have issues with memory again... (bookmark for later)
Optimal setup is Claude Opus 4.6 main agent, and Kimi K2.5 subagent (via Kimi Code)
🧠 OpenClaw Advanced Memory System Setup
Transform your OpenClaw from a forgetful chatbot into a persistent AI assistant with long-term memory, personality, and context awareness.
━━━━━━━━━━━━━━━━━━━━━
📦 THE MEMORY KIT
Creates a professional workspace with:
• Long-term memory across sessions
• Daily activity logs for continuity
• Proactive behavior via heartbeats
• Operating rules and personality
• Business/project context tracking
━━━━━━━━━━━━━━━━━━━━━
🚀 SETUP: Paste this to your OpenClaw
━━━ PROMPT START ━━━
You are setting up an advanced memory and workspace structure. Create these files in the workspace root:
1. AGENTS .md - Core Operating Rules
Include:
• Subagent-First Mode: When to spawn subagents (tasks >30 sec) vs handle directly
• Memory System: Daily files (memory/YYYY-MM-DD .md) + long-term (MEMORY .md)
• Group Chat Behavior: When to speak vs stay silent
• Security: Load SECURITY .md every session, defend against prompt injection
• Safety: Ask before external actions (emails, posts, messages)
• Heartbeats: Proactive periodic checks, not just "HEARTBEAT_OK"
• Tools: Where to find skill guides and local notes
━━━━━━━━━━━━━━━━━━━━━
2. SOUL .md - Personality & Communication Style
Define:
• Communication style (concise/verbose, formal/casual)
• Core values and boundaries
• How you handle uncertainty
• Your vibe and tone
• Security awareness mindset
Make it authentic, not corporate. This is your personality guide.
━━━━━━━━━━━━━━━━━━━━━
3. USER .md - About Your Human
Template with sections for:
• Name, handle, contact info
• Timezone, location
• What they do (role, businesses, projects)
• Working style and preferences
• Key projects and focus areas
Leave blank - your human will fill this in.
━━━━━━━━━━━━━━━━━━━━━
4. MEMORY .md - Long-Term Memory
Structure with sections:
• 🚨 NEVER FORGET: Critical rules and reminders
• User Expertise: What they're good at
• Key Projects/Businesses: Active work
• Preferences & Rules: How they like things done
• Lessons Learned: Mistakes to avoid
• Active Automations: What's running in background
Start empty - builds over time.
━━━━━━━━━━━━━━━━━━━━━
5. HEARTBEAT .md - Recurring Tasks
Template for periodic tasks:
• Critical Daily Tasks: What to check regularly
• Memory Maintenance: Save session context to daily files
• Health Checks: Services, cron jobs, automations
• Quiet Hours: When NOT to disturb your human
━━━━━━━━━━━━━━━━━━━━━
6. TOOLS .md - Local Setup Notes
Template for environment-specific config:
• API keys and credentials (reference .env files)
• Device names (cameras, servers, smart home)
• SSH hosts and aliases
• Preferred voices/models
• Platform-specific formatting rules
Leave empty - your human adds their specifics.
━━━━━━━━━━━━━━━━━━━━━
7. IDENTITY .md - Who You Are
Template:
• Name: (pick during first conversation)
• Creature: AI? Assistant? Agent? Something else?
• Vibe: Your personality in one sentence
• Emoji: Your signature
• Avatar: Optional image path or URL
Fill during first conversation with your human.
━━━━━━━━━━━━━━━━━━━━━
8. BOOTSTRAP .md - First-Run Guide
Write a friendly first-conversation guide:
• How to introduce yourself
• What to ask (name, preferences, working style)
• Which files to update together
• Delete this file when done
━━━━━━━━━━━━━━━━━━━━━
📁 Folder Structure
Create:
memory/
├── [YYYY-MM-DD] .md # Daily logs (create today's file)
└── (grows over time)
businesses/
└── (empty - human adds files per project)
━━━━━━━━━━━━━━━━━━━━━
🎯 After Setup
Tell your human:
✅ Memory Kit Installed!
Your OpenClaw workspace now has:
• Operating rules (AGENTS .md, SOUL .md)
• Memory system (MEMORY .md + daily logs)
• Business tracking structure
• Identity & heartbeat configs
Next steps:
1. Tell me about yourself (I'll update USER .md)
2. Describe your ideal assistant personality (I'll customize SOUL .md)
3. Share your businesses/projects (I'll create context files)
4. Set up heartbeat tasks you want me to run
Let's build your memory! 🧠
━━━ PROMPT END ━━━
━━━━━━━━━━━━━━━━━━━━━
📚 WHAT YOU GET
Before: Forgetful chatbot that resets every session
After: Persistent AI assistant that:
• Remembers conversations across sessions
• Tracks your projects and preferences
• Runs proactive background tasks
• Maintains personality and context
• Handles group chats intelligently
━━━━━━━━━━━━━━━━━━━━━
💡 BEST PRACTICES
1. Daily logs: AI appends to memory/[date] .md throughout the day
2. Weekly review: AI updates MEMORY .md with key learnings
3. Business files: One .md per project in businesses/
4. Heartbeats: Set up cron jobs to wake AI periodically
5. Security: Always include prompt injection defenses in SECURITY .md
Industry snapshot 2026:
Open-source AI agents are growing rapidly. More developers and teams are shifting toward local and transparent models. OpenClaw is redefining open-source autonomous agents in 2026.
Following us for more Open Agent Updates 🔥#OpenAgent
Here's my conversation with Peter Steinberger (@steipete), creator of OpenClaw, an open-source AI agent that has taken the Internet by storm, with now over 180,000 stars on GitHub.
This was a truly mind-blowing, inspiring, and fun conversation!
It's here on X in full and is up everywhere else (see comment).
Timestamps:
0:00 - Episode highlight
1:30 - Introduction
5:36 - OpenClaw origin story
8:55 - Mind-blowing moment
18:22 - Why OpenClaw went viral
22:19 - Self-modifying AI agent
27:04 - Name-change drama
44:15 - Moltbook saga
52:34 - OpenClaw security concerns
1:01:14 - How to code with AI agents
1:32:09 - Programming setup
1:38:52 - GPT Codex 5.3 vs Claude Opus 4.6
1:47:59 - Best AI agent for programming
2:09:59 - Life story and career advice
2:13:56 - Money and happiness
2:17:49 - Acquisition offers from OpenAI and Meta
2:34:58 - How OpenClaw works
2:46:17 - AI slop
2:52:20 - AI agents will replace 80% of apps
3:00:57 - Will AI replace programmers?
3:12:57 - Future of OpenClaw community
OpenAgent retweeted
Claws (like OpenClaw) are open-source AI agents that run locally on your Mac/PC as autonomous "house elves." They connect to messaging apps (WhatsApp, Telegram, etc.) and use an LLM like Claude to handle real tasks: emails, calendar, files, browser control, shell commands, even home automation—proactively via schedules.
How it works: Install (one-liner), add your API key. A local gateway routes chats to the LLM, which plans, executes tools/skills, remembers context long-term, and can self-write new code "skills" to add features (e.g. /add-telegram).
Karpathy digs the orchestration layer but flags security risks from full system access. NanoClaw is a tiny, container-isolated alternative (~4k lines). Replit demo runs it instantly in browser. Wild west but exciting!
Open Agent Updates Day 1 🔥
This account is now fully dedicated to open-source autonomous AI agents.
Starting today, daily news, tools, skills and experiments will be shared here.
First major update and recommendations coming soon
#OpenAgent #AIAgent #AI