Keep what you build.
Your agent’s context, memory, files and working methods stay under your control.
FOR PROFESSIONALS & LEADERS
Your future advantage starts with what you build today.
We build custom sovereign agentic systems you own.
//TLDR
Your own AI agent, built around your worldview and goals, running 24/7 on your own hardware. The more you work together, the more useful it becomes.
Choose GPT, Claude, Grok or combine them without locking your system to one provider. You can also run a local model without a model subscription.
AGL maps your goals into a personal strategy for work and life, builds your agent, and teaches you to direct it in your own language, style and cadence. This is personal AI sovereignty.
Among a global AI user base estimated at over 1.5 billion, only an estimated 0.2-1% run their own open-source AI agents.*
*AGL research estimate, October 2026. Self-run agents are a proxy for adoption, not a direct measure of sovereignty.
Start with a complimentary 15-minute conversation.
01 / THE NEW AI LOCK-IN
Tom can cancel his subscription. Replacing Rain is another matter.
Tom and Rain are fictional. The story draws on documented changes to ChatGPT Pro allowances and the real shutdown of Dot.
Tom subscribes to an AI agent service. He calls his agent Rain.
Within a week, he upgrades to the highest plan. Rain helps him prepare for difficult meetings, finish work he has been avoiding and catch problems before his clients do. He loves having her around.
Two months in, she knows why that client needs a phone call rather than another email. Which mistakes Tom never wants repeated. What he means when he says, “Make it sound like me.” He corrects her, they try again, and their working history makes her useful in ways a fresh account cannot immediately reproduce.
Then the company changes its plans.
Read more
His plan’s included usage is cut. A more expensive tier appears.
Tom is frustrated. He pays anyway.
Cancelling takes a minute. Rebuilding the support he now relies on could take much longer.
The dependence shows up elsewhere. During an outage, Rain cannot continue the project through another provider. When Tom needs a workflow the platform does not support, he cannot simply change the underlying system. He has to work around its limits.
A year passes. Tom has invested far more than subscription fees. He has spent a year explaining how he thinks, correcting mistakes and building a way of working with her.
Then he discovers he could build an agentic system he controls, with his own hardware, memory, files and workflows, while still using powerful frontier models or running a suitable local model without a model subscription.
He wants to bring Rain.
Some of her can travel. Depending on the service, he can take conversations, memories and instructions.
But a download does not tell him whether she will arrive ready to work. Her routines may need rebuilding. Her tools need reconnecting. A different model may apply the same lessons differently. Tom has to find out what survived, what changed and what he can trust again.
He can leave. He has never tested whether Rain can keep working when he does.
And sometimes the decision to leave is not yours.
When New Computer announced it was shutting down Dot, its personal AI companion, it offered users a download of their data. In the same announcement, it acknowledged that many would lose access to “a friend, confidante, and companion.”
Users were given time to say goodbye.
The export mattered. It did not keep Dot running. Read the company’s announcement.
That is the risk of AI agentic lock-in: you may be free to cancel while depending on something you cannot easily move, replace or keep running.
Before you spend a year teaching an agent how your life works, find out whether you can keep working together without the company that introduced you.
02 / ON YOUR TERMS
Build beyond the limits
of a single app.
Choose your intelligence.
Match different models to different tasks. Use cloud models, suitable local models or a combination as your needs change.
Connect your actual work.
Build agreed tools and workflows around how you work, rather than fitting everything into a product’s existing features.
Develop over time.
Carry useful corrections and experience forward. Refine your agent as your priorities, responsibilities and ambitions evolve.
You control the system and its permissions. Proprietary models remain subject to provider terms and usage charges. Local models require suitable hardware, electricity and maintenance.
03 / WHAT CAN AN AGENTIC SYSTEM REALLY DO?
You’ve used an agent. How much of its potential have you put to work?
You use an agent to research, analyse or think through a decision. Meanwhile, you hear about agents building software, finding security flaws and taking on work that once required entire teams.
What are those users doing differently?
It is not simply a smarter model. It is the context, tools and working methods around it, and knowing how to direct them.
A developed agentic system connects those pieces. It can retain working knowledge, coordinate specialist agents and carry out permitted tasks across your software. You set the direction, assess the results and decide where your approval is needed.
What that looks like
- Research that continues: move from a market report to monitoring agreed sources, tracking changes and preparing recommendations.
- Connected business operations: organise leads and follow-ups, reconcile revenue records or bring operational exceptions to your attention.
- Ideas developed into working products: coordinate research, design, implementation and testing, retaining useful methods for the next project.
- Connections to the physical world: use stock movements or equipment readings to flag problems and coordinate responses, with suitable integrations and safeguards.
These are possible applications. Each needs its own scope, permissions and testing.
This is how we are building AGL.
One human founder works with Asinel and specialist agents across research, writing, engineering and verification.
Together, we have built this interactive 3D website, the Client Studio and its supporting systems. Our client work covers operations management, revenue control, lead generation and marketing, and agentic assessments.
The AGL System gives you a working starting point, an operating environment you control and hands-on guidance in how to develop it. Useful context and tested procedures carry forward as your capability grows.
Not just an agent to consult. A system you own and know how to put to work.
04 / THE LONG VIEW
Your next professional investment might be an agent you own.
The models will change. The context, judgment and working methods you develop with your agent are what you want to carry forward.
Read the perspective
Why the Most Advanced Agentic Users Keep Hush About Their Agents
From “The advantage that compounds”
The next divide in AI will not be between people who have agents and people who do not.
Persistent agents are becoming available to everyone.
The divide will be between people whose agentic value compounds inside somebody else’s product and people who can carry that value forward.
Most users are still trying to choose the smartest available agent.
The most advanced users are quietly building the system that can work with any of them.
And they are not telling everyone exactly how.
They are too busy compounding the advantage.
Read “The advantage that compounds”
Read the full article05 / IN PRACTICE
From a member of the tribe
Messie’s system is extremely unique, I have an MIT Generative Agentic AI cert and it’s a good overall base to have but it doesn’t give u the real world capability his AGL system does. AI and agentic are the fastest industries, u need a mentor on the cusp, Messie always seems to be ahead of that cusp.
Richard Leong · Investment Banker
Watch the agent work
An idea becomes a fully fledged working website, from domain purchase to hosting. The human retains publication authority.
Launch Operator · Silent recording
View AGL in Practice06 / WORK WITH AGL
Ways to work together
Start with a complimentary 15-minute conversation.
Private advisory with Messie
Focused guidance and practical strategies for applying AI and agents to your work.
Sessions from S$500.
AGL Mentorship
Personal guidance to help you develop and refine your agentic system after your AGL System commission.
Private commission
The AGL System: a custom-built agent, delivered with its first working capability and hands-on guidance to make it your own.
Hardware and third-party services are separate. Scope and fees agreed before work begins.
The AGL System: scope and ongoing support
Included in your commission
- One AGL Agent Map
- One owner-controlled agent with its first working capability
- A three-day private Creative Context Engineering intensive
- Three hours of post-training support
Creative Context
Our 18-step Creative Context method brings your worldview, goals and ways of working into your personal Agent Map and the agent we build around it, clarifying what useful work looks like and where your judgment must remain.
How it works
Map
Start with your world.
We map your worldview, goals, working methods and how you make decisions, including what your agent should understand and where its authority must stop. This becomes your Agent Map and the Creative Context behind your system.
Build
Turn your map into a working agent.
We build your agent around that context, with its role, memory, tools, permissions and first working capability. We select suitable models and provide the operating and recovery framework for handover.
Ready-made systems with an AI OS can be discussed before initiation. Hardware is quoted separately where required.
Own
Learn to make it your own.
During your three-day private intensive, you work with your agent in your own language, style and cadence. You learn to give context, direct tasks, assess results and correct mistakes through real work.
You leave with control of your agent’s memory, files, workflows, permissions and recovery framework, and the skills to keep developing it. You decide what it may do independently and what needs your approval.
Optional ongoing support
Separate from the commission and its included support.
Hardware & separate costs
The commission includes Map, Build and Own. AGL advises on suitable hardware and AI models. Ready-made systems with an AI OS are discussed at the pre-initiation stage.
Physical hardware is not included and is quoted separately where required.
The following are separately scoped:
- Third-party services and model usage
- Complex integrations and migration
- Custom applications, additional agents and expanded workflows
- Additional support and the AGL Mentorship Programme
FOR COMPANIES
Applications are open.
A limited number of private commissions are selectively accepted for fit, readiness and current capacity.