This isn't some edge case. Anyone letting AI agents touch production systems has seen something like it — or will.

I use OpenClaw daily to manage the SONJJ Ecosystem — SmailPro, Smser, YChecker, and several other products. Each one has its own database, its own API, real data from real users. A Replit-style disaster isn't a matter of "if" — it's "when," unless you put guardrails in place.

After a few months of trial and error, I landed on 2 simple patterns that make working with OpenClaw significantly safer — no fancy frameworks required:

1. Keep scripts separate from cron — let the agent write script files, don't let it control cron directly
2. Use n8n webhooks as a middle layer — the agent only calls webhooks, never touches your database or product APIs

Why You Need a Middle Layer

OWASP ranks "Excessive Agency" (LLM06:2025) among the top risks when deploying AI agents — broken down into 3 dimensions:

• Excessive Functionality — giving the agent more tools than it actually needs
• Excessive Autonomy — letting the agent perform critical actions without approval
• Excessive Permissions — granting higher access than required (write access when read would do)

Sounds abstract. But map it to real OpenClaw usage and you'll recognize these immediately:

Inline cron: You ask the agent to set up a cron job. It writes a command straight into crontab -e. When something breaks, you have no idea what cron is running, can't test it in isolation, and there's no version control. The agent can schedule anything on your machine.

Custom API endpoints for the agent: You add new routes to your product's codebase — more code to write, maintain, and deploy. Over time, the codebase gets cluttered with endpoints that only serve the AI agent, not your actual users.

Direct SQL queries: This is the most dangerous one. Give the agent a connection string and it can SELECT, UPDATE, or DELETE from any table. The Replit incident from the intro? Exactly this scenario.

According to Gartner, 80% of organizations report their AI agents misbehaving, leaking data, or hallucinating. OpenClaw itself had CVE-2026-25253 (CVSS 8.8) — a vulnerability allowing remote code execution that bypassed container isolation.

The problem isn't that AI agents are unreliable. The problem is we're giving them too much direct access. The fix is simple: add a middle layer.

Pattern 1: Keep Scripts Separate From Cron

The Problem

When you ask OpenClaw to "set up a cron job to back up the database every day at 2 AM," here's what the agent does — it opens crontab -e and writes:

0 2 * * * mysqldump -u root -p'password' mydb > /tmp/backup.sql

Looks fine. Until you need to debug it. Cron runs at 2 AM while you're asleep. Next morning, no backup — and no clue why. No logs, no error output, no way to test it without waiting for the next 2 AM cycle. And since the command lives inside crontab, there's no version control — you don't know what the agent changed, or when.

This is an old DevOps problem: inline cron commands are an anti-pattern. But when an AI agent is the one writing cron entries, it gets worse — you didn't even write that command.

The Fix: Let the Agent Write Script Files

Instead of letting OpenClaw write directly into crontab, change how you give instructions:

"Write me a script backup-smailpro-db.sh to back up the SmailPro database. Don't set up cron yet — just the script."

The workflow becomes:

1. Agent writes a script file (.sh or .py) — with a shebang line, error handling, and a descriptive name
2. You review the script — read through it, check for anything unexpected
3. Test manually — run bash backup-smailpro-db.sh in your terminal, check the output
4. Add it to cron — only after the script works: 0 2 * * * /path/to/backup-smailpro-db.sh

Cron now does exactly one thing: schedule. All the logic lives in the script file — separation of concerns.

Why This Works

• Human-in-the-loop: The script is an artifact you review before it runs. The agent writes, you approve — exactly what OWASP recommends.
• Easy debugging: Something broke? Run the script in your terminal, see the output right away. No waiting for the next cron cycle.
• Version control: .sh files go into git. You know exactly what the agent changed through commit history.
• Reusable: Once written, the script can be called from anywhere — cron, a webhook, manually, or even from n8n (more on that next).
• Meaningful names: check-smailpro-health.sh, sync-user-stats.py — you know what a script does just by reading its name, instead of parsing a cryptic one-liner in crontab.

This isn't a new idea — separating scripts from cron has been best practice forever. What's new is that when an AI agent writes the script, this pattern becomes a natural safety layer: you always get a chance to review before anything runs on a schedule.

Pattern 2: n8n Webhooks — A "Firewall" for Your AI Agent

Two Bad Options

When OpenClaw needs to interact with your products — pulling metrics, checking statuses, processing data — you typically face two choices:

Build a custom API endpoint for the agent: Add a new route to your product codebase, write a controller, implement the logic, redeploy. Every new task the agent needs means more code in your product. After a few months, the codebase is littered with endpoints that serve nobody but the AI agent.

Let the agent query SQL directly: Hand OpenClaw a connection string and let it write its own queries. Fast? Sure. But the agent has full database access — SELECT * is the mild case, DROP TABLE is the nightmare. No logs, no controls, no way to know what the agent queried until something goes wrong.

The Fix: n8n Webhooks as a Middle Layer

Instead of either option, I create n8n webhooks for each specific task. The flow is straightforward:

1. Create a webhook in n8n — one webhook = one task (e.g., "count today's new SmailPro users")
2. n8n handles the backend — queries the database, calls internal APIs, transforms data
3. Returns a JSON response — the agent gets results, knows nothing about the DB or logic behind it

OpenClaw knows exactly one thing: the webhook URL. It calls that URL, gets JSON back. Done. It doesn't know where the database is, which tables exist, or what queries run.

Why n8n Instead of a Custom API?

• No code in your product: Build workflows with n8n's visual editor — drag and drop nodes. Not a single line of code touches your product codebase.
• Built-in execution logs: Every webhook call is recorded — timestamp, input payload, output, execution time. Debugging is faster than with any custom API.
• Kill switch: Agent acting up? Go into n8n, disable the webhook. Instant revocation. No redeployment, no code changes.
• Test vs Production URLs: n8n automatically creates two URLs per webhook — test freely, only activate the production URL when you're confident.
• Authentication: Supports header auth and basic auth — only requests with valid credentials get processed.
• No codebase bloat: Workflows live in n8n, completely separate from your product code. Delete a workflow and it's gone — no dead code left behind.

A Real Example

I needed OpenClaw to report daily SmailPro signups. Instead of giving the agent database access:

• Created an n8n webhook: get-new-users-today
• Workflow: Webhook node → MySQL node (query SELECT COUNT(*) FROM users WHERE DATE(created_at) = CURDATE()) → Respond to Webhook node (returns {"new_users": 42})
• OpenClaw just calls curl https://n8n.example.com/webhook/get-new-users-today → gets the number

The agent never sees the SQL query, doesn't know the table is called users, has no idea where the database lives. All it knows: call this URL, get a number.

The Common Thread

These two patterns look different but follow the same principle: always put a middle layer between your AI agent and your systems.

A script file sits between the agent and cron — you review it before it runs. An n8n webhook sits between the agent and your database — the agent only knows a URL, nothing behind it.

The more autonomy you give an agent, the more you need to restrict its direct access. Not because AI agents are dumb or untrustworthy — but because even humans need guardrails when working with production systems. AI agents are no different, they just need different guardrails.

You don't need a complex framework or an enterprise security platform to get started. A .sh file that gets reviewed before cron calls it, or an n8n webhook standing between the agent and your database — that's enough to cut your risk significantly.

If you're using OpenClaw or any AI agent, try one of these patterns. Start with the simplest task you have — you'll see the difference.

In 2003, I walked into my first internet café in Vietnam. Everything was new — the giant CRT monitor, the dial-up modem screaming through the connection, the first game loading on screen while I sat there unable to blink. The internet felt like magic. I didn't understand how it worked. I just knew it opened a world completely unlike anything I'd ever seen.

Twenty-some years later, in April 2024, I clicked "Pay" on my first $20 Claude Pro subscription. And that feeling came rushing back — identical. The same curiosity. The same sense of standing in front of something much bigger than me that I didn't fully understand yet. The same question: "What else can this thing do?"

I've been making a living online for over a decade. A generalist dev — not deep in any one thing, just enough of everything to get the job done. Code, SEO, content, running a bunch of my own websites and APIs. But when I started using AI every day — 4 to 6 hours, consistently, for nearly 2 years — I realized I wasn't just adding a new tool. I was rebuilding how I work from the ground up.

This isn't a tutorial. This isn't a tool review. This is my actual story — from that first $20 Claude Pro subscription, through the messy phase of trying everything, to finding the four tools that finally made everything click.

When Google Stopped Being My First Move

I'd been coding for fun and then for money for over a decade. I was good at Googling. I knew how to phrase exact queries, filter StackOverflow by highest votes, scan through 5-6 tabs and stitch together a solution. That was survival skill for any dev before 2024.

Then Claude changed everything.

The biggest difference wasn't speed or accuracy — it was the nature of the interaction itself. Google is search: I type keywords, read results, piece things together on my own. AI is dialogue: I describe a problem, it asks follow-up questions, I explain more, we dig deeper together. Not a smarter search engine. A completely different way of working.

And I threw everything at it:

• Code — from writing small functions to reviewing complex logic, refactoring whole modules
• SEO — keyword analysis, content strategy, technical SEO audits
• Content — brainstorming, outlining, drafting, tone adjustments
• Offpage — using AI to refine link building strategy, analyze competitors

Four to six hours a day. Not because I forced myself. Because every time I opened Claude, I'd find something else it could do that used to cost me hours of solo effort. That excitement felt exactly like the early days of discovering the internet — every day, a new corner to explore.

I basically stopped using StackOverflow. Google only comes out when I need a specific link or have to check official documentation. Everything else goes through AI.

The Phase Nobody Talks About

Then things got chaotic.

I discovered n8n — a workflow automation platform. Brilliant. I built my first workflow: auto-check email, auto-call an API. Then built another. Then deleted the first one because it no longer fit. Then rebuilt from scratch because I'd thought of a better approach. Each workflow solved one isolated problem, connected to nothing else, belonging to no system.

Meanwhile, GitHub and the AI community kept shipping new tools. Cursor. Windsurf. Claude Desktop. Cline. Continue. Aider. A new name every week, each with an impressive demo, each with an appealing use case. I'd install one, not get fluent before the next one dropped. Not get deep on one workflow before being pulled into another.

No framework for filtering. Every tool seemed good. Every workflow seemed right. But nothing connected to anything else.

I think a lot of people are in this phase right now — and it's the phase most AI workflow posts skip entirely. They jump straight from "I started using AI" to "here's my complete system," skipping over the middle part — where you're drowning in choices, tool-hopping constantly, and don't know when to stop and actually build something.

A 2026 survey found that 46% of product teams say the biggest barrier to AI adoption isn't the lack of tools — it's the lack of integration between tools and existing workflows. That number didn't surprise me. I'd lived through it.

Four Tools, One System

Things started clicking when I stopped asking "which tool is best?" and started asking "how do these tools connect?" Not because I found the perfect tool — but because I found how four tools could support each other as a single system.

n8n — The Middleware Layer

n8n is workflow automation — built to automate, connect apps, run on schedule. That's its original purpose, and it does it well. But for me, n8n plays an additional role: the middleware layer between AI and the server. Not because n8n was designed for this — but because when I looked at my workflow, it fit perfectly. Instead of letting AI call directly into a database or API server, every request goes through n8n first — processed, filtered, then returning exactly the data needed. Safer, more token-efficient, and fully under my control.

Craft Agents — The Workshop

Craft Agents is where I do the actual work every day. Building skills, creating data source connections, writing plans, testing cron jobs — anything that requires craft happens here. The chat and build experience in Craft Agents is noticeably better than any tool I've used. It's open source, so I forked it and added features to let it share skills and sources with OpenClaw. Two separate tools became one shared ecosystem.

Obsidian — The Central Brain

Obsidian isn't a note-taking app. For me, it's the knowledge management architecture for the entire system. Every document, knowledge base, project structure, AI output — all of it lives in the Obsidian vault. Both Craft Agents and OpenClaw share the same Obsidian workspace. With tags and folder structure, everything is easy to find and easy to scale.

OpenClaw — The Factory

If Craft Agents is the workshop — where you build and prepare — OpenClaw is the factory — where things run. Skills that have been tested, cron jobs that have been set up, queries that need fast results — all of that runs on OpenClaw. It doesn't replace Craft Agents. They have different roles and complement each other completely.

These four tools aren't just used separately and called a "system." They're intentionally connected — Craft Agents and OpenClaw share skills and sources through the fork, Obsidian is the shared vault for both, and n8n sits in the middle as a safe bridge to the server and external data.

Why These Four

The question I get most: "Why not just use one tool for everything?"

Because no single tool does everything well. Each one solves a different problem, and the way they support each other creates something bigger than the sum of its parts.

Why not use OpenClaw for everything? OpenClaw executes well — stable cron jobs, fast lookups, clean output. But the chat and build experience doesn't come close to Craft Agents. When you need to build a new skill, test a new source, or plan something complex — Craft Agents is smoother by a mile. These two don't compete. One builds, one executes.

Why Obsidian instead of Notion or Google Docs? Plain text and offline-first. Markdown files sit right on the machine — AI agents access them directly through the file system, no separate API calls needed, no dependency on any cloud service. When you're managing a bunch of websites and APIs, having everything in one vault with clear folder structure and tags isn't just convenient — it's essential. Obsidian is the soul of the system, not because it's trendy, but because its architecture fits perfectly with how AI agents work.

Why use n8n as middleware? n8n wasn't built for this — it's a workflow automation tool, plain and simple. But after enough trial and error, the lesson was clear: letting AI call directly into an API server is fast but wasteful (responses carry too much extra data) and carries higher security risk. I looked at what I already had and saw that n8n could stand in the middle — receive requests from AI, process them, return exactly the data needed. Safer and more token-efficient. Not an advertised feature — just the way I shaped it to fit my workflow.

Why fork Craft Agents? It's open source — anyone can fork and customize. And the instinct of someone who earns a living online is: integrate rather than reinvent. A tool that's close to what I need but missing one key feature — I don't switch tools, I fix it. Adding the ability to share skills and sources with OpenClaw turned two separate tools into one ecosystem. Don't wait for someone else to build the bridge. Build it yourself.

The problem a lot of teams face isn't a shortage of AI tools — it's the missing connections between them. This four-tool stack solves that — not with a super-tool, but by wiring the right tools together.

What's Coming Next

This post is just the big picture — the story of the journey, not a setup guide. But behind this four-tool stack are a lot of specifics I'm gradually documenting:

• How to fork Craft Agents and add skill/source sharing with OpenClaw — step by step, from zero
• Obsidian vault architecture for AI agents — how I organize folders, tags, and naming conventions so both Craft Agents and OpenClaw can read everything
• Building n8n as an API middleware — designing safe workflows between AI and a production server
• My actual daily workflow — from receiving a task to getting it done, where AI steps in, and which tool handles which part

None of this is generic tutorial content. These are things I'm actually using, documented from real operations — with enough context for you to adapt to your own work, not copy wholesale.

If you want those posts as soon as they're out, subscribe to the newsletter below. No spam, no sales. Just posts when I actually have something worth sharing.

Still the Same Curiosity

Nearly two years. From that first $20 Claude Pro to a system managing a whole stack of websites and APIs I've been building for over a decade. But the numbers aren't the point. The biggest shift was in thinking — from tool user to system builder. From someone searching for answers to someone designing how to ask questions.

AI doesn't replace skills. It amplifies them. Everyone has access to AI — but actually putting it to work in a meaningful way is something else entirely. Making AI tools interact and fit together is harder still. Research says AI coding assistants boost productivity 20-30% — but that number only means something when you have a system to take advantage of it. Using AI piecemeal, differently every day, keeps you stuck in experiment mode. With a system, AI actually becomes part of how you work.

If you're in the chaotic phase right now — tool-hopping, deleting workflows, not knowing where to stop — that's normal. I was there too. Everyone goes through it before things start to click.

You don't need to start with four tools at once. Start with one. Get deep on it. Then add the second when you feel the gap. Everyone's journey is different — what matters is building with intention, not building because of FOMO.

Sitting in front of Claude now feels the same as sitting in front of that CRT monitor in the internet café back in 2003. Still the same curiosity. The difference is — this time, I know what I'm building.

1. Introduction to SonJJ's credit system

SonJJ's credit system is the main payment unit for all API requests. This is one of the most outstanding features of SonJJ compared to other API services in the market:

• Credits never expire - You can buy and use them anytime, without worrying about usage time limits
• Charges only on success - Credits are only counted when the request returns status code 200
• Fair cost - Each type of endpoint has a different cost, reflecting the actual value it provides
• Attractive bonuses - Save up to 35% when buying larger credit packages

"Credits do not expire. And it comes in cheap, if you buy larger plans."

2. Simple guide to creating an API Key

You just need to log in to the system management area through this link https://my.sonjj.com/. Then you only need to select the menu item named API

You will see a button called "Change API KEY", click on it and you will get an API Key to use.

3. Guide to buying credits from wallet

Before buying credits, you need to ensure you have added money to your SonJJ wallet. If not, you can refer to the guide on adding funds to your SonJJ wallet.

Available credit packages

Step-by-step credit purchase process

1. Log in to your SonJJ account
2. Click on the "Credit" tab in the navigation bar
3. In the "Buy Credits (does not expire)" section, choose the credit package that fits your needs
4. Click on the "Buy Now" button
5. The system will automatically pay from your SonJJ wallet (through "Pay from your wallet")
6. Credits will be added to your account immediately

Saving tip: If you plan to use the API for a long time, choosing Plan-3 with a 35% bonus will help you save significant costs. With 133,650 credits (including bonus), you can make millions of requests to low-cost endpoints.

4. Credit calculation mechanism for each API request

SonJJ applies a smart credit calculation mechanism: only counting credits for successful requests (returning status_code=200). This ensures you only pay when you actually receive value from the API.

"Cost of each successful request (status_code=200)"

Credit costs are allocated based on the value and complexity of each endpoint:

Explaining cost differences

There are clear differences in costs between endpoints:

• Simple endpoints (0.05 credits): Inbox query endpoints have the lowest cost due to simple processing
• Standard endpoints (1-2 credits): Message, list, and create endpoints have average costs
• Complex endpoints (>10 credits): Endpoints like check_index have the highest cost (14 credits) due to their high value and complex processing

Special note: Costs will vary between plan packages. The above pricing applies to PLAN-1. With higher packages that include bonuses, the actual cost for each request will be lower.

5. Benefits of SonJJ's credit system

SonJJ's credit system brings many benefits to users:

Fair cost

You only pay based on the actual value received. Simple endpoints have low costs, while endpoints providing higher value will have correspondingly higher costs. This ensures a balance between cost and benefit.

Easy budget forecasting

With fixed credit costs for each type of request, you can easily estimate the API budget for your project. For example, if you know you will make 1,000 requests to the inbox endpoint daily, you will spend 50 credits/day (1,500 credits/month).

Savings with scale

The more credits you buy, the more you save. With the 35% bonus from Plan-3, the actual cost for each request will decrease significantly. This is a major advantage for businesses using the API with high frequency.

No worries about deadlines

Credits never expire, allowing you to buy in large quantities to take advantage of bonuses without worrying about having to use them all within a specific time period.

6. API cost optimization strategies

To optimize costs when using SonJJ's API, you can apply the following strategies:

Choose low-cost endpoints when possible

For example, if you only need to check if there are new emails, use the /inbox endpoint (0.05 credits) instead of /message (1-2 credits). Only use expensive endpoints when absolutely necessary.

Focus on credit packages with high bonuses

Plan-3 with a 35% bonus will help you save significant costs in the long run. Instead of buying Plan-1 multiple times, consider investing in Plan-3 from the start.

Optimize API call frequency

Instead of calling the API continuously, set up reasonable intervals between calls. This not only saves credits but also reduces load on your system.

Cost optimization example:

A company needs to check 10,000 emails monthly:

• Option 1: Using Single API /v1/check_index (14 credits/request) = 140,000 credits/month
• Option 2: Combining inbox checks (0.05 credits) and only using message checks (1 credit) when needed = ~20,000 credits/month

→ Savings: 120,000 credits/month (equivalent to over $120 with Plan-1)

7. Frequently asked questions

8. Conclusion and next steps

SonJJ's credit system provides a fair and efficient approach to API usage. With features like never-expiring credits, charging only on success, and attractive bonuses for large packages, SonJJ helps you optimize costs significantly.

To start using the API and take advantage of the credit system:

1. Add funds to your SonJJ wallet (if you haven't already)
2. Purchase an appropriate credit package (recommend Plan-3 if you have long-term usage plans)
3. Create an API key and start integrating the API into your application
4. Apply the cost optimization strategies mentioned

Start today and enjoy the benefits of SonJJ's flexible, cost-effective API system!

English has truly been my weak point, but with this decision, I must change that. I've decided to systematically improve my English skills, not just for better communication but to connect more effectively with the international community. This will enable me to share more projects, case studies, and practical experiences with users worldwide.

My goal is clear: to bring optimization knowledge and solutions to more people.

Hello there, I don't know where you're coming from, but if you've made it here today and are reading this article, then we truly have a connection. Perhaps our common goal is the same - to learn methods and optimize time when working on computers. And that's exactly why I built this website.

Sonjj.com is a website specialized in providing API tools or solutions to help save as much computer work time as possible... Its mission is to help online workers, programmers, marketers, or anyone facing tons of repetitive daily tasks - reclaim their time. If you already understand the purpose of this website, then don't waste any more time - let's get started! In the next section, I will guide you step by step on how to register for a SonJJ.com account - the first step to begin your journey of reclaiming your valuable time.

II. Sign up for sonjj account

Imagine this as creating a key to open the door to our ecosystem. Once you have this key, you can move freely between different "rooms" (websites within the system) without having to knock on the door from the beginning.

Access my.sonjj.com

First, open your browser and access my.sonjj.com - this is the control center for our entire ecosystem. To be honest, I designed it similar to how Google operates their accounts. Sign in once, use everywhere.

The interesting thing is that when you try to log in to any website in the ecosystem such as smailpro.com, smser.net, or cardgener.com, the system will automatically direct you to my.sonjj.com for authentication and then redirect you back. No need to remember multiple different passwords - simplifying everything!

Fill in registration information

You have three ways to create a sonjj account:

Method 1: Manual Registration (traditional form)

This method may be a bit "classic" but gives you complete control:

1. Enter your name
2. Fill in your email correctly (this will be your primary ID to recover your account if needed)
3. Create a password (the more complex, the better)
4. Confirm your password (just to ensure you didn't mistype) After submitting the form, check your email for confirmation. Seriously, don't skip this step - it's how the system knows your email actually exists.

Method 2: Quick Registration with Google

If you're a fan of "optimizing every second" like me, this option is for you:

1. Click on the "Sign in with Google" button
2. Select the Google account you want to use
3. Confirm access permissions And done! The system automatically retrieves your email and name from your Google account. No need for email verification because Google has already done that.

Method 3: Use an existing sonjj.com account

If you already have a sonjj.com account:

1. Click on the "Sign in with SonJJ" option
2. Enter your sonjj.com account information

Email Verification

With manual registration, you'll receive a confirmation email within minutes. The email contains an activation link - click on it and your account will be activated immediately.

The good thing here is that if you don't receive the email, there's a "Resend confirmation email" button right on the page. No need to fill out the form again from the beginning - I hate having to re-enter information.

First Login

After verifying your email or registering via Google/SonJJ, you'll be redirected to the main dashboard. This isn't a complicated admin page with graphs and charts that overwhelm you - I've designed it simple enough for newcomers to easily navigate. You can see the image below; if you see an interface like that, it means you've successfully created an account and logged in.

III. Congratulations! You Have Completed the First Step

So you've successfully created your SonJJ.com account and become familiar with the admin interface. Easy, wasn't it? This is just the first step to begin your journey of automating your daily tasks.

To truly harness the power of these APIs, the next step you need to take is to deposit money into your account and get an API key to start using them.

In the next article, I will guide you in detail on how to deposit money into your SonJJ account and obtain an API key. Every minute you invest in this initial setup will help you save hours later on.

See you in the article "Guide to Depositing Money Into Your Account And Converting Money to Points".

1. Usage History

• Automatically saves created profiles in the local browser memory
• Limits storage to a maximum of 50 most recent profiles
• Easy access and reuse of information

2. Copy All

• Allows copying all created profile information with just one click
• Saves time and reduces manual operations for users

3. Custom Views

• Provides templates to display information according to specific needs
• Example: Google view will only display fields necessary for creating a Google account
• Will continue to add more templates in the future

4. Custom Domain

• Allows users to change the domain of the created email address
• Flexibility in creating email addresses according to needs

These new features are designed to enhance the user experience, increase usability, and expand the customization capabilities of the tool.

The new interface is designed to be more intuitive and polished, helping users easily monitor and select the type of email they need to create. Additionally, we have added a Gmail list from the 5th server and will continue to update in the future. Visit here to try the new tool: https://smailpro.com/temporary-email