Meta-Schema
The programmable meaning layer for enterprise agents.
Most organizations do not lack data.
They lack agreement about what the data means.
The database has a field. The API has a payload. The spreadsheet has a column. The policy has a definition. The agent has a tool description.
They almost line up.
Almost is where operational risk lives.
A name changes in one system. A status means three different things. A required field is required only sometimes. An agent is handed a schema with no memory of the policy behind it.
Then everyone starts asking the same question:
what does this mean here?
Meta-Schema starts from a simple assumption:
the schema should carry the meaning, not merely the shape.
The Primitive#
A Meta-Schema is a schema that describes the shape, meaning, constraints, provenance, interfaces, and allowed use of something in a system.
It is not only a validator.
It is the shared object of interpretation.
Define the concept once. Let people, applications, databases, documents, workflows, and agents read the same agreement.
Definition#
A Meta-Schema is a durable declaration of what something is, how it may change, where it came from, what it connects to, and how software may safely use it.
That is the sentence to memorize.
It is intentionally plain.
Because the hard part of a schema is rarely the syntax. The hard part is getting a company, a product, and a set of systems to mean the same thing at the same time.
Properties#
A Meta-Schema carries the properties ordinary schemas usually leave outside the boundary:
- shape: the fields, types, cases, and relationships
- meaning: the business definition and human language around the concept
- constraints: what must be true for the object to be valid
- provenance: where the value came from and why it should be trusted
- interfaces: how applications, agents, documents, and APIs may read or write it
- history: how the concept changed over time
- authority: who can define, approve, or override it
- examples: concrete cases that make the definition testable
The goal is not to make schemas heavier.
The goal is to stop pretending the rest of this information is optional.
What It Is Not#
A Meta-Schema is not just JSON Schema, GraphQL, SQL DDL, OpenAPI, a data catalog, or a glossary.
Those are useful surfaces.
A Meta-Schema is the source object those surfaces should be derived from.
The glossary explains it. The validator checks it. The database stores it. The API exposes it. The agent calls it.
The Meta-Schema is the agreement they share.
Example#
This is not only a data model.
It is also policy context, workflow input, agent guidance, documentation, validation, and integration structure.
The important part is not that it can generate artifacts.
The important part is that the artifacts agree because they came from one declaration.
Why Now#
Agents make weak schemas expensive.
Humans can ask a follow-up question. Software usually cannot. Agents can, but only if the system gives them something better than a field list and a few hopeful descriptions.
An enterprise agent needs to know:
- what an object is
- what words mean in context
- what actions are allowed
- what constraints apply
- what source is authoritative
- what changed recently
- what uncertainty remains
That is a schema problem.
But it is not solved by more syntax.
It is solved by treating schemas as living institutional agreements.
Hosted#
Meta-Schema is the hosted place to define, inspect, publish, and govern Meta-Schemas at meta-schema.com.
The product is for teams that need software and people to share language: product teams, operations teams, compliance teams, data teams, and agent builders.
You can start with a simple concept.
Name it. Define it. Add constraints. Attach examples. Connect it to the systems that use it.
Over time, the organization gets a property most systems never quite achieve:
it knows what its own words mean.
Relationship To Meta-Effects#
Meta-Schema defines shared meaning.
Meta-Effects executes declarations.
They are related, but they are not the same primitive.
A Meta-Schema tells the system what a concept is. A Meta-Effect tells the system what can happen with that concept.
One names the world.
The other makes the world executable.