Back to Directory/Other

Avrotize

Schema conversion and schema-driven code generation MCP server.

OtherPythonv3.4.3
View on GitHub

Avrotize & Structurize

mcp-name: io.github.clemensv/avrotize

PyPI version Python Versions Build Status License: MIT Downloads

📚 Documentation & Examples | 🎨 Conversion Gallery

Avrotize is a "Rosetta Stone" for data structure definitions, allowing you to convert between numerous data and database schema formats and to generate code for different programming languages.

It is, for instance, a well-documented and predictable converter and code generator for data structures originally defined in JSON Schema (of arbitrary complexity).

The tool leans on the Apache Avro-derived Avrotize Schema as its schema model.

  • Programming languages: Python, C#, Java, TypeScript, JavaScript, Rust, Go, C++
  • SQL Databases: MySQL, MariaDB, PostgreSQL, SQL Server, Oracle, SQLite, BigQuery, Snowflake, Redshift, DB2
  • Other databases: KQL/Kusto, SurrealDB, MongoDB, Cassandra, Redis, Elasticsearch, DynamoDB, CosmosDB
  • Data schema formats: Avro, JSON Schema, JSON Structure, XML Schema (XSD), Protocol Buffers 2 and 3, ASN.1, Apache Parquet, JSON Type Definition (JTD), CDDL, CUE, FlatBuffers, Apache Thrift IDL, Smithy IDL, Cap'n Proto, RAML 1.0 Data Types, and OpenAPI 3.x

Installation

You can install Avrotize from PyPI, having installed Python 3.10 or later:

pip install avrotize

For MCP server support (avrotize mcp), install with the MCP extra:

pip install "avrotize[mcp]"

For SQL database support (sql2a command), install the optional database drivers:

# PostgreSQL
pip install avrotize[postgres]

# MySQL
pip install avrotize[mysql]

# SQL Server
pip install avrotize[sqlserver]

# All SQL databases
pip install avrotize[all-sql]

Usage

Avrotize provides several commands for converting schema formats via Avrotize Schema.

Converting to Avrotize Schema:

Converting from Avrotize Schema:

Converting to and from JSON Structure:

Inferring schemas from data:

Generate code from Avrotize Schema:

Generate code from JSON Structure:

Other commands:

  • avrotize validate - Validate JSON instances against Avro or JSON Structure schemas.
  • avrotize mcp - Run Avrotize as a local MCP server exposing conversion tools to MCP clients.
  • avrotize pcf - Create the Parsing Canonical Form (PCF) of an Avrotize Schema.
  • avrotize validate-tmsl - Validate TMSL scripts locally against documented object structure.

Overview

MCP server

You can run Avrotize as a local MCP server over stdio:

avrotize mcp

Catalog-ready metadata files are included:

To publish to the official MCP Registry:

mcp-publisher validate server.json
mcp-publisher publish server.json

The MCP server exposes tools to:

  • describe server capabilities and routing guidance (describe_capabilities)
  • list available conversion commands (list_conversions)
  • inspect a conversion command (get_conversion)
  • execute conversions (run_conversion)

You can use Avrotize to convert between Avro/Avrotize Schema and other schema formats like JSON Schema, XML Schema (XSD), Protocol Buffers (Protobuf), ASN.1, and database schema formats like Kusto Data Table Definition (KQL) and SQL Table Definition. That means you can also convert from JSON Schema to Protobuf going via Avrotize Schema.

You can also generate C#, Java, TypeScript, JavaScript, and Python code from Avrotize Schema documents. The difference to the native Avro tools is that Avrotize can emit data classes without Avro library dependencies and, optionally, with annotations for JSON serialization libraries like Jackson or System.Text.Json.

The tool does not convert data (instances of schemas), only the data structure definitions.

Mind that the primary objective of the tool is the conversion of schemas that describe data structures used in applications, databases, and message systems. While the project's internal tests do cover a lot of ground, it is nevertheless not a primary goal of the tool to convert every complex document schema like those used for devops pipeline or system configuration files.

Why?

Data structure definitions are an essential part of data exchange, serialization, and storage. They define the shape and type of data, and they are foundational for tooling and libraries for working with the data. Nearly all data schema languages are coupled to a specific data exchange or storage format, locking the definitions to that format.

Avrotize is designed as a tool to "unlock" data definitions from JSON Schema or XML Schema and make them usable in other contexts. The intent is also to lay a foundation for transcoding data from one format to another, by translating the schema definitions as accurately as possible into the schema model of the target format's schema. The transcoding of the data itself requires separate tools that are beyond the scope of this project.

The use of the term "data structure definition" and not "data object definition" is quite intentional. The focus of the tool is on data structures that can be used for messaging and eventing payloads, for data serialization, and for database tables, with the goal that those structures can be mapped cleanly from and to common programming language types.

Therefore, Avrotize intentionally ignores common techniques to model object-oriented inheritance. For instance, when converting from JSON Schema, all content from allOf expressions is merged into a single record type rather than trying to model the inheritance tree in Avro.

Avrotize Schema

Avrotize Schema is a schema model that is a full superset of the popular Apache Avro Schema model. Avrotize Schema is the "pivot point" for this tool. All schemas are converted from and to Avrotize Schema.

Since Avrotize Schema is a superset of Avro Schema and uses its extensibility features, every Avrotize Schema is also a valid Avro Schema and vice versa.

Why did we pick Avro Schema as the foundational schema model?

Avro Schema ...

  • provides a simple, clean, and concise way to define data structures. It is quite easy to understand and use.
  • is self-contained by design without having or requiring external references. Avro Schema can express complex data structure hierarchies spanning multiple namespace boundaries all in a single file, which neither JSON Schema nor XML Schema nor Protobuf can do.
  • can be resolved by code generators and other tools "top-down" since it enforces dependencies to be ordered such that no forward-referencing occurs.
  • emerged out of the Apache Hadoop ecosystem and is widely used for serialization and storage of data and for data exchange between systems.
  • supports native and logical types that cover the needs of many business and technical use cases.
  • can describe the popular JSON data encoding very well and in a way that always maps cleanly to a wide range of programming languages and systems. In contrast, it's quite easy to inadvertently define a JSON Schema that is very difficult to map to a programming language structure.
  • is itself expressed as JSON. That makes it easy to parse and generate, which is not the case for Protobuf or ASN.1, which require bespoke parsers.

It needs to be noted here that while Avro Schema is great for defining data structures, and data classes generated from Avro Schema using this tool or other tools can be used to with the most popular JSON serialization libraries, the Apache Avro project's own JSON encoding has fairly grave interoperability issues with common usage of JSON. Avrotize defines an alternate JSON encoding

in avrojson.md.

Avro Schema does not support all the bells and whistles of XML Schema or JSON Schema, but that is a feature, not a bug, as it ensures the portability of the schemas across different systems and infrastructures. Specifically, Avro Schema does not support many of the data validation features found in JSON Schema or XML Schema. There are no pattern, format, minimum, maximum, or required keywords in Avro Schema, and Avro does not support conditional validation.

In a system where data originates as XML or JSON described by a validating XML Schema or JSON Schema, the assumption we make here is that data will be validated using its native schema language first, and then the Avro Schema will be used for transformation or transfer or storage.

Adding CloudEvents columns for database tables

When converting Avrotize Schema to Kusto Data Table Definition (KQL), SQL Table Definition, or Parquet Schema, the tool can add special columns for CloudEvents attributes. CNCF CloudEvents is a specification for describing event data in a common way.

The rationale for adding such columns to database tables is that messages and events commonly separate event metadata from the payload data, while that information is merged when events are projected into a database. The metadata often carries important context information about the event that is not contained in the payload itself. Therefore, the tool can add those columns to the database tables for easy alignment of the message context with the payload when building event stores.

Convert Proto schema to Avrotize Schema

avrotize p2a <path_to_proto_file> [--out <path_to_avro_schema_file>]

Parameters:

  • <path_to_proto_file>: The path to the Protobuf schema file to be converted. If omitted, the file is read from stdin.
  • --out: The path to the Avrotize Schema file to write the conversion result to. If omitted, the output is directed to stdout.

Conversion notes:

  • Proto 2 and Proto 3 syntax are supported.
  • Proto package names are mapped to Avro namespaces. The tool does resolve imports and consolidates all imported types into a single Avrotize Schema file.
  • The tool embeds all 'well-known' Protobuf 3.0 types in Avro format and injects them as needed when the respective types are imported. Only the Timestamp type is mapped to the Avro logical type 'timestamp-millis'. The rest of the well-known Protobuf types are kept as Avro record types with the same field names and types.
  • Protobuf allows any scalar type as key in a map, Avro does not. When converting from Proto to Avro, the type information for the map keys is ignored.
  • The field numbers in message types are not mapped to the positions of the fields in Avro records. The fields in Avro are ordered as they appear in the Proto schema. Consequently, the Avrotize Schema also ignores the extensions and reserved keywords in the Proto schema.
  • The optional keyword results in an Avro field being nullable (union with the null type), while the required keyword results in a non-nullable field. The repeated keyword results in an Avro field being an array of the field type.
  • The oneof keyword in Proto is mapped to an Avro union type.
  • All options in the Proto schema are ignored.

Convert Avrotize Schema to Proto schema

avrotize a2p <path_to_avro_schema_file> [--out <path_to_proto_directory>] [--naming <naming_mode>] [--allow-optional]

Parameters:

  • <path_to_avro_schema_file>: The path to the Avrotize Schema file to be converted. If omitted, the file is read from stdin.
  • --out: The path to the Protobuf schema directory to write the conversion result to. If omitted, the output is directed to stdout.
  • --naming: (optional) Type naming convention. Choices are snake, camel, pascal.
  • --allow-optional: (optional) Enable support for 'optional' fields.

Conversion notes:

  • Avro namespaces are resolved into distinct proto package definitions. The tool will create a new .proto file with the package definition and an import statement for each namespace found in the Avrotize Schema.
  • Avro type unions [] are converted to oneof expressions in Proto. Avro allows for maps and arrays in the type union, whereas Proto only supports scalar types and message type references. The tool will therefore emit message types containing a single array or map field for any such case and add it to the containing type, and will also recursively resolve further unions in the array and map values.
  • The sequence of fields in a message follows the sequence of fields in the Avro record. When type unions need to be resolved into oneof expressions, the alternative fields need to be assigned field numbers, which will shift the field numbers for any subsequent fields.

Convert CUE schema subset to Avrotize Schema

avrotize cue2a <path_to_cue_file> [--out <path_to_avro_schema_file>] [--namespace <avro_schema_namespace>]

### Convert FlatBuffers schema to Avrotize Schema

```bash
avrotize fbs2a <path_to_fbs_file> [--out <path_to_avro_schema_file>] [--namespace <avro_schema_namespace>]

### Convert Apache Thrift IDL to Avrotize Schema

```bash
avrotize thrift2a <path_to_thrift_file> [--out <path_to_avro_schema_file>] [--namespace <avro_schema_namespace>]

### Convert Smithy IDL data shapes to Avrotize Schema

```bash
avrotize smithy2a <path_to_smithy_file> [--out <path_to_avro_schema_file>] [--namespace <avro_schema_namespace>]

### Convert Cap'n Proto schema to Avrotize Schema

```bash
avrotize capnp2a <path_to_capnp_file> [--out <path_to_avro_schema_file>] [--namespace <avro_schema_namespace>]

### Convert RAML Data Types to Avrotize Schema

```bash
avrotize raml2a <path_to_raml_file> [--out <path_to_avro_schema_file>] [--namespace <avro_schema_namespace>]

Parameters:

  • <path_to_cue_file>: The path to the CUE file to be converted. If omitted, the file is read from stdin.
  • --out: The path to the Avrotize Schema file to write. If omitted, the output is directed to stdout.
  • --namespace: (optional) Namespace for generated Avrotize Schema records. If omitted, a simple CUE package name is used as the namespace.

Supported subset / limitations:

  • Supported: schema-style definitions #Name: { ... }, top-level fields as a generated record, required fields name: T, optional fields name?: T, primitive types string, int, float, number, bool, bytes, and null (int maps to Avro long; number maps to Avro double).
  • Supported: nested structs as records, open lists [...T] as arrays, open structs { [string]: T } as maps, references to other definitions (#Other), disjunctions as Avro unions, nullable disjunctions such as *null | T or T | null, string-literal disjunctions as Avro enums with sanitized symbols, and simple defaults like name: T | *default.
  • Out of scope: full CUE constraint solving, imports beyond a simple package declaration, complex constraints and comprehensions, computed or unified expressions, bounds and regex constraints (>0, =~"..."), interpolation, and package/module resolution. Unsupported constructs are skipped or mapped to a broad string type with a conversion note rather than causing a crash.

Example:

package demo

#Person: {
  name: string
  age?: int
  tags: [...string]
  status: "new" | "active"
}

Convert Avrotize Schema to CUE schema subset

avrotize a2cue <path_to_avro_schema_file> [--out <path_to_cue_file>] [--namespace <cue_package_hint>]

Conversion notes:

  • Avro records become CUE definitions (#Name: { ... }), fields become name: Type, nullable unions become optional fields where possible, enums become string disjunctions, arrays become [...T], and maps become { [string]: T }.
  • Avro int and long both emit as CUE int; float emits as float; double emits as number. Avro namespaces are reduced to a simple CUE package name using the last namespace segment.
  • The converter emits the same practical schema subset accepted by cue2a; it does not attempt to reconstruct CUE constraints, imports, comprehensions, or computed expressions.

Convert CUE schema subset to JSON Structure

avrotize cue2s <path_to_cue_file> [--out <path_to_structure_file>] [--namespace <namespace>]

cue2s bridges through an intermediate Avrotize Schema file and therefore uses the same supported CUE subset and limitations as cue2a.

Convert JSON Structure to CUE schema subset

avrotize s2cue <path_to_structure_file> [--out <path_to_cue_file>] [--namespace <cue_package_hint>]

s2cue bridges through an intermediate Avrotize Schema file and emits the same practical CUE schema subset as a2cue.

  • <path_to_fbs_file>: The path to the FlatBuffers .fbs schema file to be converted. If omitted, the file is read from stdin.
  • --out: The path to the Avrotize Schema file to write the conversion result to. If omitted, the output is directed to stdout.
  • --namespace: (optional) Override the FlatBuffers namespace for the Avrotize Schema.

Conversion notes and limitations:

  • FlatBuffers namespace declarations are mapped to Avro namespaces. table and struct declarations are mapped to Avro records; FlatBuffers structs carry fixed inline layout semantics that Avro does not represent, so this is documented on the generated record.
  • FlatBuffers enums are mapped to Avro enums. Integer enum values are preserved in the non-standard Avrotize ordinals annotation; consumers that only understand Avro enum symbols may ignore those integer values.
  • FlatBuffers unions are mapped to Avro unions of their member record types.
  • Scalar mappings: signed 8/16/32-bit integers map to Avro int; unsigned 32-bit integers and all 64-bit integers map to Avro long; uint64/ulong values beyond signed 64-bit range cannot be represented exactly by Avro long; float and double map to Avro float and double; string maps to Avro string.
  • Vectors map to Avro arrays, except [ubyte]/[uint8], which maps to Avro bytes.
  • Table fields without (required) are emitted as nullable Avro fields with null defaults. FlatBuffers field defaults are preserved as the Avrotize fbsDefault annotation for nullable fields.
  • root_type is preserved as a record-level root_type annotation.

Convert Avrotize Schema to FlatBuffers schema

avrotize a2fbs <path_to_avro_schema_file> [--out <path_to_fbs_file>] [--namespace <flatbuffers_namespace>]

- `<path_to_thrift_file>`: The path to the Apache Thrift IDL file to be converted. If omitted, the file is read from stdin.
- `--out`: The path to the Avrotize Schema file to write the conversion result to. If omitted, the output is directed to stdout.
- `--namespace`: (optional) Override the Avro namespace. Without an override, `namespace *` is used, then the first language-specific namespace.

Conversion notes and limitations:

- `struct`, `exception`, and `union` declarations are emitted as Avro records. Thrift unions are modeled as records whose fields are all nullable; Avro does not enforce the Thrift rule that at most one field is set.
- `enum` declarations are emitted as Avro enums. Ordinals are preserved in an `ordinals` annotation. Symbols and names that are not valid Avro names are sanitized.
- `typedef` aliases are resolved to their target type. `const` declarations and `service` definitions are skipped because they do not describe persistent data structures.
- `set<T>` is emitted as an Avro array and does not preserve uniqueness semantics. `map<string,V>` is emitted as an Avro map; maps with non-string keys are emitted as arrays of `{ key, value }` records.
- `include` statements are recorded by the parser but are not recursively resolved by the converter. Convert included IDL files separately or pre-expand them before conversion.

### Convert Avrotize Schema to Apache Thrift IDL

```bash
avrotize a2thrift <path_to_avro_schema_file> [--out <path_to_thrift_file>] [--namespace <thrift_namespace>]

- `<path_to_smithy_file>`: The path to the Smithy 2.0 IDL file to be converted. If omitted, the file is read from stdin.
- `--out`: The path to the Avrotize Schema file to write the conversion result to. If omitted, the output is directed to stdout.
- `--namespace`: (optional) Override the Smithy `namespace` for the emitted Avro namespace.

Conversion notes:

- Phase 1 supports Smithy data shapes only: `structure`, `union`, `enum`, `intEnum`, `list`, `map`, and scalar shape references. `service`, `operation`, `resource`, HTTP/protocol traits, mixins beyond simple parsing, and `apply` statements are explicitly out of scope and are skipped without failing conversion.
- Smithy `namespace` maps to Avro `namespace`. `@documentation` maps to Avro `doc`; `@required` makes a field non-nullable; non-required fields are nullable and default to `null`; `@default` maps to the Avro field default where Avro permits it. `@deprecated` and `@tags` are carried into `doc` text.
- Smithy `union` shapes are represented as Avro records whose alternatives are nullable fields, preserving member names while keeping the Avro schema valid. Avro field unions convert back to Smithy `union` shapes.
- Smithy `intEnum` converts to an Avro enum with an `ordinals` annotation; Avro itself stores enum symbols, so integer values are metadata for round-tripping.
- Smithy `list` and `map` members map to Avro arrays and maps. Avro maps are string-keyed, so Smithy map keys should be `String`; other key declarations are noted but cannot be represented as Avro map keys.
- Scalar mappings include `Blob`→`bytes`, `Boolean`→`boolean`, `String`→`string`, integer widths→`int`/`long`, floats→`float`/`double`, `Timestamp`→valid Avro `long` with `timestamp-millis`, `BigInteger`→`string`, `BigDecimal`→`double`, and `Document`→`string`.

### Convert Avrotize Schema to Smithy IDL data shapes

```bash
avrotize a2smithy <path_to_avro_schema_file> [--out <path_to_smithy_file>] [--namespace <smithy_namespace>]

- `<path_to_capnp_file>`: The path to the Cap'n Proto `.capnp` schema file to be converted. If omitted, the file is read from stdin.
- `--out`: The path to the Avrotize Schema file to write the conversion result to. If omitted, the output is directed to stdout.
- `--namespace`: Optional Avro namespace. If omitted, the input file name is used.

Conversion notes and limitations:

- Structs map to Avro records, enums map to Avro enums, and field ordinals are stored as `capnpOrdinal` metadata. Enum ordinals are stored as `capnpOrdinals`.
- Cap'n Proto fields are pointer-default/zero-default optional in practice; Avrotize emits nullable Avro fields (`["null", T]`) with `default: null` for non-`Void` fields.
- Primitive mapping: `Bool`→`boolean`; `Int8`/`Int16`/`Int32`→`int`; `Int64`→`long`; unsigned integers including `UInt64`→`long` (range checks are not represented); `Float32`→`float`; `Float64`→`double`; `Text`→`string`; `Data`→`bytes`; `Void`→`null`; `List(T)`→Avro array. No invalid Avro logical types are emitted.
- Anonymous and named Cap'n Proto unions are represented as nullable fields tagged with `capnpUnion` metadata rather than as exclusive Avro unions; exclusivity constraints are not enforced by Avro.
- Groups are represented as inline nested Avro records. Nested structs and enums are emitted as named Avro types in derived nested namespaces.
- The file id is accepted but not used as an Avro namespace seed. `interface`, `const`, `annotation`, imports, and using declarations are skipped.

### Convert Avrotize Schema to Cap'n Proto schema

```bash
avrotize a2capnp <path_to_avro_schema_file> [--out <path_to_capnp_file>] [--namespace <namespace_note>]

- `<path_to_raml_file>`: The path to the RAML 1.0 file or library. If omitted, the file is read from stdin.
- `--out`: The path to the Avrotize Schema file to write. If omitted, the output is directed to stdout.
- `--namespace`: (optional) Namespace for generated Avro named types.

Conversion notes:

- Phase 1 supports RAML 1.0 **Data Types** in the `types:` section only. API resources, methods, traits, resourceTypes, securitySchemes, annotations, and external `!include` expansion are explicitly out of scope and are ignored or left as inert YAML values.
- Object types with `properties:` become Avro records. Optional properties (`name?` or `required: false`) become nullable Avro fields with `null` first and default `null`.
- Scalar mappings are `string`→`string`, `number`→`double`, `integer`→`long`, `boolean`→`boolean`, `file`→`bytes`, and `nil`→`null`.
- RAML date/time precision is normalized to valid Avro logical types: `date-only`→`int`/`date`, `time-only`→`int`/`time-millis`, and `datetime-only`/`datetime`→`long`/`timestamp-millis`.
- Arrays use `T[]` or `type: array` with `items`. Unions use `A | B`; nullable `T?` is treated as `nil | T`.
- Maps use the documented RAML convention `properties: { "//": T }` or `additionalProperties: T` and become Avro maps.
- RAML enums become Avro enums; symbols are sanitized to Avro identifiers. Non-string enum base types are emitted as Avro enum symbols, so literal value typing is not preserved.

### Convert Avrotize Schema to RAML Data Types

```bash
avrotize a2raml <path_to_avro_schema_file> [--out <path_to_raml_file>] [--namespace <namespace_to_strip>]

Parameters:

  • <path_to_avro_schema_file>: The path to the Avrotize Schema file to be converted. If omitted, the file is read from stdin.
  • --out: The path to the FlatBuffers .fbs file to write the conversion result to. If omitted, the output is directed to stdout.
  • --namespace: (optional) Override the FlatBuffers namespace.

Conversion notes and limitations:

  • Avro records are emitted as FlatBuffers tables, Avro enums as FlatBuffers enums, arrays as vectors, and Avro bytes as [ubyte].
  • Nullable Avro unions (["null", T]) become optional FlatBuffers fields. Multi-branch Avro unions of named types are emitted as FlatBuffers union declarations.
  • Avro namespaces map to FlatBuffers namespaces. The top record, or the record annotated with root_type, is emitted as root_type.
  • Avro maps, logical types, fixed types, aliases, validation annotations, and arbitrary custom annotations have no direct FlatBuffers equivalent and are either simplified to compatible field types or omitted.

Convert Avrotize Schema to ASN.1 schema

avrotize a2asn <path_to_avro_schema_file> [--out <path_to_asn1_file>] [--module <asn1_module_name>]

Parameters:

  • <path_to_avro_schema_file>: The path to the Avrotize Schema file to be converted. If omitted, the file is read from stdin.
  • --out: The path to the ASN.1 module file to write the conversion result to. If omitted, the output is directed to stdout. The ASN.1 module name is derived from the output file name when --module is not given.
  • --module: (optional) Override the ASN.1 module name.

Conversion notes and limitations:

  • The emitted module uses DEFINITIONS AUTOMATIC TAGS, which lets the ASN.1 compiler disambiguate OPTIONAL and CHOICE tags automatically.
  • Avro records map to SEQUENCE, Avro enums to ENUMERATED, arrays to SEQUENCE OF, and fixed to OCTET STRING (SIZE(n)).
  • Avro maps map to SEQUENCE OF SEQUENCE { key UTF8String, value ... }.
  • Nullable unions (["null", T]) become OPTIONAL members; multi-branch unions become an ASN.1 CHOICE.
  • Logical types map to ASN.1 useful types: date → DATE, time-millis/time-micros → TIME-OF-DAY, timestamp-* → DATE-TIME; decimal, uuid, and duration are represented as REAL/UTF8String because ASN.1 has no exact equivalents.
  • Identifiers are sanitized to strict X.680 (hyphens instead of underscores); the module is validated by round-tripping through the asn1tools compiler.

Convert JSON Structure Schema to ASN.1 schema

avrotize s2asn <path_to_structure_schema_file> [--out <path_to_asn1_file>] [--module <asn1_module_name>]

Parameters:

  • <path_to_structure_schema_file>: The path to the JSON Structure schema file to be converted. If omitted, the file is read from stdin.
  • --out: The path to the ASN.1 module file to write the conversion result to. If omitted, the output is directed to stdout. The ASN.1 module name is derived from the output file name when --module is not given.
  • --module: (optional) Override the ASN.1 module name.

Conversion notes and limitations:

  • This is a direct JSON Structure → ASN.1 mapping; it does not route through Avrotize Schema, so the richer JSON Structure type system is preserved.
  • Sized integer types map to INTEGER with faithful range constraints (for example int8 → INTEGER (-128..127), uint32 → INTEGER (0..4294967295)).
  • object maps to SEQUENCE, array to SEQUENCE OF, set to SET OF, map to SEQUENCE OF SEQUENCE { key UTF8String, value ... }, tuple to a positional SEQUENCE, and choice to CHOICE.
  • String enum/const values map to ENUMERATED (preserving the symbol labels and their ordinals); integer enum/const values map to an INTEGER (v1 | v2 | ...) value constraint (preserving the exact values).
  • $extends merges the base object's properties into the derived SEQUENCE; canonical {"type": {"$ref": ...}} references, bare {"$ref": ...} references, and nested definition namespaces are all resolved.
  • Non-required or nullable members become OPTIONAL. Types without an exact ASN.1 equivalent (uuid, uri, duration, jsonpointer) are represented as UTF8String; any maps to the ASN.1 open type ANY.

Convert FlatBuffers schema to JSON Structure

avrotize fbs2s <path_to_fbs_file> [--out <path_to_json_structure_file>] [--namespace <avro_schema_namespace>]

Parameters:

  • <path_to_fbs_file>: The path to the FlatBuffers .fbs schema file to be converted. If omitted, the file is read from stdin.
  • --out: The path to the JSON Structure file to write the conversion result to. If omitted, the output is directed to stdout.
  • --namespace: (optional) Override the namespace used by the Avrotize Schema bridge.

Conversion notes:

  • This conversion bridges through Avrotize Schema (fbs2a followed by a2s) and therefore shares the FlatBuffers-to-Avro mapping limitations listed above.
  • The FlatBuffers root_type record is used as the JSON Structure root when present.

Convert JSON Structure to FlatBuffers schema

avrotize s2fbs <path_to_json_structure_file> [--out <path_to_fbs_file>] [--namespace <flatbuffers_namespace>]

Parameters:

  • <path_to_json_structure_file>: The path to the JSON Structure schema file to be converted. If omitted, the file is read from stdin.
  • --out: The path to the FlatBuffers .fbs file to write the conversion result to. If omitted, the output is directed to stdout.
  • --namespace: (optional) Override the FlatBuffers namespace.

Conversion notes:

  • This conversion bridges through Avrotize Schema (s2a followed by a2fbs) and therefore shares the Avro-to-FlatBuffers mapping limitations listed above.

  • JSON Structure constraints and metadata that do not survive the Avrotize Schema bridge are not represented in the generated FlatBuffers schema.

  • --out: The path to the Thrift IDL file to write. If omitted, the output is directed to stdout.

  • --namespace: (optional) Override the emitted namespace * value.

Conversion notes and limitations:

  • Avro records are emitted as Thrift struct definitions with sequential field ids. Avro enums are emitted as Thrift enums.
  • Nullable Avro unions (["null", T]) are emitted as optional fields; other multi-branch Avro unions are approximated as string.
  • Avro arrays are emitted as list<T> and Avro maps as map<string,V>. Thrift set and non-string map-key semantics cannot be recovered from Avro.
  • Avro logical types and custom annotations are not represented in Thrift IDL.

Convert Apache Thrift IDL to JSON Structure

avrotize thrift2s <path_to_thrift_file> [--out <path_to_structure_file>] [--namespace <avro_schema_namespace>]

Converts Thrift IDL to JSON Structure by first converting to Avrotize Schema. The Thrift-to-Avro limitations above therefore apply.

Convert JSON Structure to Apache Thrift IDL

avrotize s2thrift <path_to_structure_file> [--out <path_to_thrift_file>] [--namespace <thrift_namespace>]

Converts JSON Structure to Thrift IDL by first converting to Avrotize Schema. The Avro-to-Thrift limitations above therefore apply.

  • --out: The path to the Smithy IDL file to write. If omitted, the output is directed to stdout.
  • --namespace: (optional) Override the Smithy namespace to emit.

Conversion notes:

  • Avro records become Smithy structure shapes; nullable unions become optional members; non-null fields are emitted with @required.

  • Avro enums become Smithy enum shapes, or intEnum when an ordinals annotation is present. Avro arrays and maps become Smithy list and map shapes. Avro unions with multiple non-null alternatives become Smithy union shapes.

  • Service and operation modeling is explicitly out of phase-1 scope; this command emits Smithy data shapes only.

  • --out: The path to the Cap'n Proto schema file to write. If omitted, the output is directed to stdout.

  • --namespace: Optional namespace note included as a comment in the generated file.

Conversion notes and limitations:

  • Avro records map to struct, enums to enum, arrays to List(T), strings to Text, bytes/fixed to Data, and nullable unions to the non-null Cap'n Proto field type.
  • Fields are emitted with deterministic sequential ordinals (@0, @1, ...). A stable generated file id is emitted; replace it if the schema needs a project-owned Cap'n Proto id.
  • Multi-branch Avro unions are emitted as nested choice structs or union blocks when capnpUnion metadata is present. Avro maps are represented as lists of generated key/value entry structs because Cap'n Proto has no direct map primitive.

Convert Cap'n Proto schema to JSON Structure

avrotize capnp2s <path_to_capnp_file> [--out <path_to_json_structure_file>] [--namespace <avro_schema_namespace>] [--naming <naming_mode>] [--avro-encoding]

This conversion bridges through an intermediate Avrotize Schema, so the Cap'n Proto limitations documented for capnp2a apply.

Convert JSON Structure to Cap'n Proto schema

avrotize s2capnp <path_to_json_structure_file> [--out <path_to_capnp_file>] [--namespace <namespace_note>]

This conversion bridges through an intermediate Avrotize Schema, so the Avro-to-Cap'n Proto limitations documented for a2capnp apply.

  • <path_to_avro_schema_file>: The path to the Avrotize Schema file. If omitted, the file is read from stdin.
  • --out: The path to the RAML file to write. If omitted, the output is directed to stdout.
  • --namespace: (optional) Namespace to strip from generated RAML type references.

Conversion notes:

  • The converter emits a #%RAML 1.0 Library with a types: map. Full RAML API resource/method conversion is explicitly out of scope.
  • Avro records become RAML object types, nullable fields are emitted with the name? optional-property form, enums become enum:, arrays become T[], maps use properties: { "//": T }, and unions become A | B.
  • Avro logical types map back to RAML date-only, time-only, or datetime. datetime-only cannot be distinguished from Avro timestamp logical types on reverse conversion.

Convert JSON schema to Avrotize Schema

avrotize j2a <path_to_json_schema_file> [--out <path_to_avro_schema_file>] [--namespace <avro_schema_namespace>] [--split-top-level-records]

Parameters:

  • <path_to_json_schema_file>: The path to the JSON schema file to be converted. If omitted, the file is read from stdin.
  • --out: The path to the Avrotize Schema file to write the conversion result to. If omitted, the output is directed to stdout.
  • --namespace: (optional) The namespace to use in the Avrotize Schema if the JSON schema does not define a namespace.
  • --split-top-level-records: (optional) Split top-level records into separate files.

Conversion notes:

Convert Avrotize Schema to JSON schema

avrotize a2j <path_to_avro_schema_file> [--out <path_to_json_schema_file>] [--naming <naming_mode>]

Parameters:

  • <path_to_avro_schema_file>: The path to the Avrotize Schema file to be converted. If omitted, the file is read from stdin.
  • --out: The path to the JSON schema file to write the conversion result to. If omitted, the output is directed to stdout.
  • --naming: (optional) Type naming convention. Choices are snake, camel, pascal, default.

Conversion notes:

Convert XML Schema (XSD) to Avrotize Schema

avrotize x2a <path_to_xsd_file> [--out <path_to_avro_schema_file>] [--namespace <avro_schema_namespace>]

Parameters:

  • <path_to_xsd_file>: The path to the XML schema file to be converted. If omitted, the file is read from stdin.
  • --out: The path to the Avrotize Schema file to write the conversion result to. If omitted, the output is directed to stdout.
  • --namespace: (optional) The namespace to use in the Avrotize Schema if the XML s

Installation

Source-derived launch command. Check the maintainer’s required arguments and credentials before running:

bash
uvx avrotize

Set up in your AI client

Merge this template into ~/Library/Application Support/Claude/claude_desktop_config.json. Keep existing servers. Add any arguments, credentials, and permissions required by the maintainer; this template has not been install-tested.

json
{
  "mcpServers": {
    "io-github-clemensv-avrotize": {
      "command": "uvx",
      "args": [
        "avrotize"
      ]
    }
  }
}

Restart Claude Desktop completely for changes to take effect. Confirm the server appears connected in the client’s tool list, then try a read-only example from its documentation.

Claude Desktop setup reference

Package

avrotizepypi

Compatible MCP Clients

Avrotize works with any MCP-compatible client. Copy the config snippet from the Configuration section above and add it to the file shown for your client, then restart the application.

  • Claude Desktop~/Library/Application Support/Claude/claude_desktop_config.jsonRestart Claude Desktop completely for changes to take effect.
  • Cursor~/.cursor/mcp.jsonRestart Cursor for changes to take effect.
  • VS Code.vscode/mcp.jsonReload VS Code window for changes to take effect.
  • Windsurf~/.codeium/windsurf/mcp_config.jsonRestart Windsurf for changes to take effect.
  • Claude Code.mcp.jsonSave at the project root, then start Claude Code in that project and review the MCP server approval prompt. Keep real credentials out of shared files.

Learn More

MCP vs ACP: What's the Difference Between Agent Protocols?MCP connects an agent to tools and data. ACP connects an editor to an agent. They sound similar and get confused constantly, but the JetBrains-Zed backed ACP and Anthropic's MCP were built to stack on top of each other.Read articlePinterest vs Microsoft Advertising MCP ServersTwo ad platforms shipped MCP servers on the same day, one week ahead of Cannes Lions. Pinterest's is an invite-only alpha; Microsoft's is a broadly available pilot. Neither is indexed in MCPFind's directory yet. Here's the real difference between them.Read articleHome Assistant MCP Server: 4 Options Compared for ClaudeHome Assistant has no vendor-official MCP server, so four community projects fill the gap, each with a different architecture. This guide compares voska/hass-mcp, homeassistant-ai/ha-mcp, achetronic/hass-mcp, and a custom-component option so you can pick the one that matches your setup.Read article