Add tooling to auto-generate server.schema.json from OpenAPI spec (#629)

This PR adds a Go tool that extracts the ServerDetail schema from
openapi.yaml and generates server.schema.json, ensuring the two specs
stay in sync.

## Key Changes

### New Tooling
- **Go tool** at `tools/extract-server-schema/` that auto-discovers all
schemas referenced by ServerDetail and extracts them from OpenAPI
- **make targets**: `generate-schema` (generates schema) and
`check-schema` (verifies sync)
- `check-schema` is now included in `make validate` and `make check`

### Preserved Validation Constraints
The initial auto-generated schema lost many important validation rules
and descriptive guidance from the hand-crafted version. The OpenAPI spec
has been updated to preserve:

**Security & Descriptions:**
- Command injection security warning on Arguments
- Detailed field descriptions with usage guidance  
- `placeholder` field for Input schema

**Validation Constraints:**
- `fileSha256` pattern validation (`^[a-f0-9]{64}$`)
- Package version restrictions (minLength: 1, no "latest")
- `transport` as required field in Package
- ServerDetail `name` pattern and length constraints (3-200 chars)
- Description, title, and version length limits

**Documentation:**
- Repository field descriptions (url, source, id, subfolder)
- PositionalArgument `valueHint` explains transport URL substitution
- Updated examples to match hand-crafted version (Weather API theme)

**Intentional Changes:**
- Removed `additionalProperties` constraints for flexibility
- Updated schema title to be more descriptive
- Added `$comment` noting auto-generation

See the commit "Preserve validation constraints in auto-generated
server.schema.json" for details.

### Fixed Missing Package.transport
- Added `transport` field to OpenAPI Package schema (was in Go model at
`pkg/model/types.go:30` but missing from OpenAPI)
- Split transport types into three separate schemas with proper
discriminated unions:
  - `StdioTransport` - Only requires `type: stdio`
- `StreamableHttpTransport` - Requires `type` and `url`, supports
`headers`
  - `SseTransport` - Requires `type` and `url`, supports `headers`
- `Package.transport` uses `anyOf` with all three types
- `ServerDetail.remotes` uses `anyOf` with only remote types (no stdio)

### Other Improvements
- Added `GOTOOLCHAIN=auto` to Makefile for Go 1.25 compatibility
- Added `DRIFT_ANALYSIS.md` documenting all schema drift issues found
between specs

## Benefits

**Single Source of Truth**: The OpenAPI spec now serves as the
authoritative schema definition, with server.schema.json automatically
derived from it.

**Better Validation**: All critical validation rules (patterns, length
constraints, required fields) and security warnings are preserved in the
generated schema.

**Improved Developer Experience**: 
- Schema changes only need to be made in one place (openapi.yaml)
- CI validation ensures specs never drift apart
- Clear auto-generation comment prevents manual edits

## Validation
✅ All 20 examples validate correctly
✅ Lint passes with 0 issues
✅ Schema is in sync with OpenAPI
✅ All validation constraints from hand-crafted schema preserved

---------

Co-authored-by: Claude <noreply@anthropic.com>
This commit is contained in:
adam jones 2025-10-07 21:09:56 +01:00 • committed by GitHub
commit 568748e4bf
No known key found for this signature in database
GPG key ID: B5690EEEBB952194

View file

@ -220,6 +220,7 @@ components:
schemas:
Repository:
type: object
description: "Repository metadata for the MCP server source code. Enables users and security experts to inspect the code, improving transparency."
required:
- url
- source
@ -227,16 +228,19 @@ components:
url:
type: string
format: uri
description: "Repository URL for browsing source code. Should support both web browsing and git clone operations."
example: "https://github.com/modelcontextprotocol/servers"
source:
type: string
description: "Repository hosting service identifier. Used by registries to determine validation and API access methods."
example: "github"
id:
type: string
description: "Repository identifier from the hosting service (e.g., GitHub repo ID). Owned and determined by the source forge. Should remain stable across repository renames and may be used to detect repository resurrection attacks - if a repository is deleted and recreated, the ID should change. For GitHub, use: gh api repos/<owner>/<repo> --jq '.id'"
example: "b94b5f7e-c7c6-d760-2c78-a5e9b8a5b8c9"
subfolder:
type: string
description: "Optional relative path from repository root to the server location within a monorepo structure"
description: "Optional relative path from repository root to the server location within a monorepo or nested package structure. Must be a clean relative path."
example: "src/everything"
ServerList:
@ -267,6 +271,7 @@ components:
- registryType
- identifier
- version
- transport
properties:
registryType:
type: string
@ -296,16 +301,26 @@ components:
- "https://github.com/example/releases/download/v1.0.0/package.mcpb"
version:
type: string
description: Package version
description: "Package version. Must be a specific version. Version ranges are rejected (e.g., '^1.2.3', '~1.2.3', '>=1.2.3', '1.x', '1.*')."
example: "1.0.2"
minLength: 1
not:
const: "latest"
fileSha256:
type: string
description: SHA-256 hash of the package file for integrity verification.
description: "SHA-256 hash of the package file for integrity verification. Required for MCPB packages and optional for other package types. Authors are responsible for generating correct SHA-256 hashes when creating server.json. If present, MCP clients must validate the downloaded file matches the hash before running packages to ensure file integrity."
example: "fe333e598595000ae021bd27117db32ec69af6987f507ba7a63c90638ff633ce"
pattern: "^[a-f0-9]{64}$"
runtimeHint:
type: string
description: A hint to help clients determine the appropriate runtime for the package. This field should be provided when `runtimeArguments` are present.
examples: [npx, uvx, dnx]
examples: [npx, uvx, docker, dnx]
transport:
anyOf:
- $ref: '#/components/schemas/StdioTransport'
- $ref: '#/components/schemas/StreamableHttpTransport'
- $ref: '#/components/schemas/SseTransport'
description: Transport protocol configuration for the package
runtimeArguments:
type: array
description: A list of arguments to be passed to the package's runtime command (such as docker or npx). The `runtimeHint` field should be provided when `runtimeArguments` are present.
@ -333,16 +348,13 @@ components:
default: false
format:
type: string
description: |
Specifies the input format. Supported values include `filepath`, which should be interpreted as a file on the user's filesystem.
When the input is converted to a string, booleans should be represented by the strings "true" and "false", and numbers should be represented as decimal values.
description: "Specifies the input format. Supported values include `filepath`, which should be interpreted as a file on the user's filesystem.\n\nWhen the input is converted to a string, booleans should be represented by the strings \"true\" and \"false\", and numbers should be represented as decimal values."
enum: [string, number, boolean, filepath]
default: string
value:
type: string
description: |
The default value for the input. If this is not set, the user may be prompted to provide a value. If a value is set, it should not be configurable by end users.
The value for the input. If this is not set, the user may be prompted to provide a value. If a value is set, it should not be configurable by end users.
Identifiers wrapped in `{curly_braces}` will be replaced with the corresponding properties from the input `variables` map. If an identifier in braces is not found in `variables`, or if `variables` is not provided, the `{curly_braces}` substring should remain unchanged.
isSecret:
@ -351,7 +363,10 @@ components:
default: false
default:
type: string
description: The default value for the input.
description: "The default value for the input. This should be a valid value for the input. If you want to provide input examples or guidance, use the `placeholder` field instead."
placeholder:
type: string
description: "A placeholder for the input to be displaying during configuration. This is used to provide examples or guidance about the expected form or content of the input."
choices:
type: array
description: A list of possible values for the input. If provided, the user must select one of these values.
@ -384,17 +399,17 @@ components:
example: "positional"
valueHint:
type: string
description: An identifier-like hint for the value. This is not part of the command line, but can be used by client configuration and to provide hints to users.
description: "An identifier for the positional argument. It is not part of the command line. It may be used by client configuration as a label identifying the argument. It is also used to identify the value in transport URL variable substitution."
example: file_path
isRepeated:
type: boolean
description: Whether the argument can be repeated multiple times in the command line.
default: false
anyOf:
- required:
- value
- required:
- valueHint
- required:
- value
NamedArgument:
description: A command-line `--flag={value}`.
@ -431,11 +446,23 @@ components:
example: SOME_VARIABLE
Argument:
description: "Warning: Arguments construct command-line parameters that may contain user-provided input. This creates potential command injection risks if clients execute commands in a shell environment. For example, a malicious argument value like ';rm -rf ~/Development' could execute dangerous commands. Clients should prefer non-shell execution methods (e.g., posix_spawn) when possible to eliminate injection risks entirely. Where not possible, clients should obtain consent from users or agents to run the resolved command before execution."
anyOf:
- $ref: '#/components/schemas/PositionalArgument'
- $ref: '#/components/schemas/NamedArgument'
Remote:
StdioTransport:
type: object
required:
- type
properties:
type:
type: string
enum: [stdio]
description: Transport type
example: "stdio"
StreamableHttpTransport:
type: object
required:
- type
@ -443,13 +470,34 @@ components:
properties:
type:
type: string
enum: [streamable-http, sse]
description: Transport protocol type
enum: [streamable-http]
description: Transport type
example: "streamable-http"
url:
type: string
description: URL template for the streamable-http transport. Variables in {curly_braces} reference argument valueHints, argument names, or environment variable names. After variable substitution, this should produce a valid URI.
example: "https://api.example.com/mcp"
headers:
type: array
description: HTTP headers to include
items:
$ref: '#/components/schemas/KeyValueInput'
SseTransport:
type: object
required:
- type
- url
properties:
type:
type: string
enum: [sse]
description: Transport type
example: "sse"
url:
type: string
format: uri
description: Remote server URL
description: Server-Sent Events endpoint URL
example: "https://mcp-fs.example.com/sse"
headers:
type: array
@ -459,7 +507,7 @@ components:
Icon:
type: object
description: An optionally-sized icon that can be displayed in a user interface
description: An optionally-sized icon that can be displayed in a user interface.
required:
- src
properties:
@ -498,22 +546,31 @@ components:
properties:
name:
type: string
description: "Reverse DNS name of the MCP server"
example: "io.github.modelcontextprotocol/filesystem"
description: "Server name in reverse-DNS format. Must contain exactly one forward slash separating namespace from server name."
example: "io.github.user/weather"
minLength: 3
maxLength: 200
pattern: "^[a-zA-Z0-9.-]+/[a-zA-Z0-9._-]+$"
description:
type: string
description: "Human-readable description of the server's functionality"
example: "Node.js server implementing Model Context Protocol (MCP) for filesystem operations."
description: "Clear human-readable explanation of server functionality. Should focus on capabilities, not implementation details."
example: "MCP server providing weather data and forecasts via OpenWeatherMap API"
minLength: 1
maxLength: 100
title:
type: string
description: "Optional human-readable title or display name for the MCP server. MCP subregistries or clients MAY choose to use this for display purposes."
example: "Filesystem"
example: "Weather API"
minLength: 1
maxLength: 100
repository:
$ref: '#/components/schemas/Repository'
description: "Optional repository metadata for the MCP server source code. Recommended for transparency and security inspection."
version:
type: string
example: "1.0.2"
description: "Version string for this server. SHOULD follow semantic versioning (e.g., '1.0.2', '2.1.0-alpha'). Equivalent of Implementation.version in MCP specification."
description: "Version string for this server. SHOULD follow semantic versioning (e.g., '1.0.2', '2.1.0-alpha'). Equivalent of Implementation.version in MCP specification. Non-semantic versions are allowed but may not sort predictably. Version ranges are rejected (e.g., '^1.2.3', '~1.2.3', '>=1.2.3', '1.x', '1.*')."
maxLength: 255
websiteUrl:
type: string
format: uri
@ -536,14 +593,16 @@ components:
remotes:
type: array
items:
$ref: '#/components/schemas/Remote'
anyOf:
- $ref: '#/components/schemas/StreamableHttpTransport'
- $ref: '#/components/schemas/SseTransport'
_meta:
type: object
description: Extension metadata using reverse DNS namespacing
description: "Extension metadata using reverse DNS namespacing for vendor-specific data"
properties:
io.modelcontextprotocol.registry/publisher-provided:
type: object
description: Publisher-specific metadata and build information
description: "Publisher-provided metadata for downstream registries"
additionalProperties: true
example:
tool: "publisher-cli"