What is MCL?
MCL stands for Metaverse Construction Language. In Sigla, it is the layer that makes a Space readable to AI. A Space is not only a screen and not only a conversation. It has an interface, data, files, logic, and connected capabilities. MCL describes all of them in one structure.
The important property is not the syntax. It is that the code explains the application. An agent can see what the app does, where its state lives, and which actions are available. It does not need a separate specification that can drift out of date.
Why AI Native, not just AI-assisted
An AI-assisted product adds a chat interface to software built for people. An AI-native product is built so AI can read, change, run, and verify it. The difference shows up after the first version:
- Can the agent see the current implementation?
- Can it change the app in place?
- Can it run the result and check whether the change worked?
- Can a person and an agent use the same result?
MCL is designed around "yes" to those questions. That is what makes it an AI Native Framework rather than a code generator with a chat box.
What MCL expresses
| Part of the Space | What the code describes |
|---|---|
| Interface | The view, layout, components, and interaction |
| Data | Storage, records, and state |
| Logic | Actions, flows, and methods |
| Files | The files and assets the Space uses |
| Skills | The capabilities bound to the Space |
| MCP | External tools and data the Space can reach |
| Automations | Scheduled and triggered work |
Because these live in the same structure, a change to one part can be understood against the others. The agent is not editing a screenshot or guessing a database schema.
How AI builds and changes a Space
- Read. The agent inspects the current MCL view, MCS logic, data, files, and logs.
- Plan. It decides which parts of the Space need to change for the request.
- Change. It edits the same code that runs the Space.
- Run. The change takes effect in the running application.
- Verify. You check the render view; the agent can also inspect data and logs.
This loop is why the result can keep changing. It is also why you can ask the agent to operate the app: the structure it edits is the structure it can read at runtime.
MCL vs traditional app frameworks
| Traditional framework | MCL | |
|---|---|---|
| Optimized for | Humans writing code | AI reading, changing, running, and verifying |
| Code and context | Separate docs and runtime | Code is the context |
| Change loop | Edit, build, deploy, test | Edit, run, verify in the Space |
| Who operates the app | People, through the UI | People and agents |
| Deployment | Separate step | The Space is running as it is built |
This does not make traditional frameworks obsolete. MCL is for a different job: software that a person and an agent build and use together.
A short example
<view>
[research_board]
[h1] { "text": "Weekly research" }
[task_list]
{each: self.items as item}
[task_item]
[text] { "text": "$item.title" }
</view>
<script>
state:
items: []
def refresh:
@desc: "Refresh the research list"
@flow:
self.items = await mcs.fetch_items()
</script>
The exact syntax is less important than the shape: the view, state, and action are in one readable structure. An agent can change the list, the state, or the action without a separate translation step.
What MCL is not
- It is not a drag-and-drop no-code builder. You describe the result; AI writes and changes the code.
- It is not a prompt wrapper around a fixed template. The Space can change structurally.
- It is not open source yet. MCL is already used in Sigla; the open-source release is in preparation.
FAQ
Is MCL a programming language?
It is the language used to describe a Space, and it is part of Sigla's AI Native Framework. You do not have to write it yourself; AI writes and changes it from your request.
Do I need to learn MCL to use Sigla?
No. You describe the app or the change. MCL is what makes the result readable to the agent. If you want to inspect or hand-edit a Space, dev mode shows the MCL view, MCS logic, data, files, and logs.
Is MCL open source?
Not yet. MCL is already used in Sigla. The open-source release is in preparation and will be announced here.
How is MCL different from an app generator?
An app generator produces a first version and then hands you a codebase. MCL keeps the running app and its code in the same structure, so the agent can keep changing and verifying it.
Can MCL connect to external tools?
Yes. MCL includes access for Agents, Skills, and MCP. A Space can also import capabilities from MCP into its Skill library.