Summary: An MCP server for browsing, managing, and exporting Kafka Schema Registry resources across multiple registries.
Check server health/MCP compliance and discover available MCP resources and URI templates.
Inspect registries: list configured/default registries, get registry info, and test one or all connections.
Browse schemas/contexts: list subjects and contexts, fetch schemas by subject/version or global ID, list schema versions, find subjects by schema ID, and read global/subject configs and modes.
Register schemas, check compatibility, and create contexts where the edition allows writes.
Export individual schemas or full subjects, with options for config, metadata, and versions.
Count contexts, schemas, and schema versions.
Edition-dependent access: viewer for read-only browsing/export/compare, editor for registration/evolution, admin for configuration/deletion/migration/bulk operations; supports VIEWONLY safety and OAuth 2.1.
Packaged as a Docker container for easy deployment and configuration across environments, with various tags for stable and development versions
Mentioned as one of the supported OAuth providers for authentication
Provides Helm charts for simplified deployment to Kubernetes environments
Supported as an OAuth provider for authentication and authorization
Supports deployment to Kubernetes clusters via Helm charts
Supported as an OAuth provider for authentication
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@MCP Kafka Schema Reg list all schema contexts"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
A comprehensive Model Context Protocol (MCP) server that provides Claude Desktop and other MCP clients with tools for Kafka Schema Registry operations. Features advanced schema context support, multi-registry management, and comprehensive schema export capabilities.
🎯 True MCP Implementation: Uses FastMCP 4.1+ on MCP Python SDK 2 with MCP 2026-07-28 compliance. The same process negotiates
2025-11-25with older clients. Fully compatible with Claude Desktop and other MCP clients using JSON-RPC overstdio.
Latest Version: v3.1.0 | Docker: aywengo/kafka-schema-reg-mcp:stable
🎛️ v3.0 editions. The server runs as
viewer(read-only: browse, export, compare, compatibility),editor(viewer + schema registration and evolution) oradmin(full access: configuration, deletion, migration, bulk operations, disaster recovery). Plain:latest/:stableimages areviewer;-editorand-admintags and the Helmeditionvalue opt into more.SLIM_MODEis a deprecated alias foreditor. See docs/editions.md and docs/v3-migration-guide.md.
Related MCP server: mcp-confluent
# Latest stable release
docker pull aywengo/kafka-schema-reg-mcp:stable
# Viewer is the default; choose editor or admin explicitly when you need writes
docker run -e MCP_EDITION=editor -e SCHEMA_REGISTRY_URL=http://localhost:8081 aywengo/kafka-schema-reg-mcp:stable
# OR run with the full admin surface for administrators/SRE
docker run -e MCP_EDITION=admin -e SCHEMA_REGISTRY_URL=http://localhost:8081 aywengo/kafka-schema-reg-mcp:stableCopy a ready-to-use configuration from config-examples/:
# macOS
cp config-examples/claude_desktop_stable_config.json ~/Library/Application\ Support/Claude/claude_desktop_config.json
# Linux
cp config-examples/claude_desktop_stable_config.json ~/.config/claude-desktop/config.jsonRestart Claude Desktop and try these prompts:
"List all schema contexts"
"Show me the subjects in the production context"
"Register a new user schema with fields for id, name, and email"
🤖 Claude Desktop Integration - Direct MCP integration with natural language interface
🏢 Multi-Registry Support - Manage up to 8 Schema Registry instances simultaneously
📋 Schema Contexts - Logical grouping for production/staging environment isolation
🔄 Schema Migration - Cross-registry migration with backup and verification
📊 Comprehensive Export - JSON, Avro IDL formats for backup and documentation
🔒 Production Safety - VIEWONLY mode and per-registry access control
🔐 OAuth 2.1 Authentication - Enterprise-grade security with scope-based permissions
📈 Real-time Progress - Async operations with progress tracking and cancellation
🔗 Resource Linking - HATEOAS navigation with enhanced tool responses
🧪 MCP 2026-07-28 - FastMCP 4 / MCP SDK 2; one process also serves 2025-11-25 clients
🎛️ Three editions - viewer (39 tools), editor (42) and admin (67) from one codebase
📖 See detailed feature descriptions: docs/api-reference.md
This project includes 5 specialized Claude Code skills – 4 for automated schema development workflows plus 1 for schema context comparison (/context-compare, documented below):
/schema-generate - Generate production-ready Avro schemas from natural language
/schema-generate event UserRegistered "user registration with userId, email, timestamp"/schema-evolve - Safely evolve schemas with automatic compatibility checking
/schema-evolve user-profile "add optional phoneNumber field"/migration-plan - Create detailed migration plans between environments
/migration-plan development staging/lint-and-test - Automated quality assurance workflows
/lint-and-test quick # Before commit (2-3s)
/lint-and-test fix # Auto-fix issues (20-30s)
/lint-and-test pre-push # Before push (10-15s)Quick Start: Read .claude-code/SKILLS_GUIDE.md - 5-minute tutorial
Complete Reference: .claude-code/skills/README.md - Full documentation
Setup Summary: .claude-code/skills/README.md - Configuration details
✅ Natural language schema generation with templates
✅ Automatic compatibility checking (BACKWARD, FORWARD, FULL)
✅ Migration planning with rollback procedures
✅ Pre-commit and pre-push quality automation
✅ Integration with Black, Ruff, isort, Flake8
✅ Docker-based test execution
✅ Comprehensive error handling and auto-fix
Try it now: /schema-generate event TestEvent "test with id and timestamp"
# Promoted channel. `:stable` moves only when Tag as Stable runs.
docker pull aywengo/kafka-schema-reg-mcp:stable
docker pull aywengo/kafka-schema-reg-mcp:stable-editor
docker pull aywengo/kafka-schema-reg-mcp:stable-admin
# This release. Admin on the 3.1.0 tag is 67 tools.
docker pull aywengo/kafka-schema-reg-mcp:3.1.0-adminMCP_EDITION selects the tool surface at startup. Plain tags run viewer.
# Read-only browsing
docker run -e SCHEMA_REGISTRY_URL=http://localhost:8081 aywengo/kafka-schema-reg-mcp:stable
# Schema registration and evolution
docker run -e MCP_EDITION=editor -e SCHEMA_REGISTRY_URL=http://localhost:8081 aywengo/kafka-schema-reg-mcp:stable
# Full surface, including deletion and migration
docker run -e MCP_EDITION=admin -e SCHEMA_REGISTRY_URL=http://localhost:8081 aywengo/kafka-schema-reg-mcp:stable⚠️
SLIM_MODEis deprecated. It is an alias forMCP_EDITION=editorand will be removed in 4.0.0. Do not set both variables.See docs/editions.md for the complete per-edition tool list and docs/v3-migration-guide.md for the upgrade path.
git clone https://github.com/aywengo/kafka-schema-reg-mcp
cd kafka-schema-reg-mcp
pip install -r requirements.txt
python kafka_schema_registry_unified_mcp.pydocker-compose up -d # Includes Schema Registry for testing📖 Detailed installation guide: docs/deployment.md
export SCHEMA_REGISTRY_URL="http://localhost:8081"
export SCHEMA_REGISTRY_USER="" # Optional
export SCHEMA_REGISTRY_PASSWORD="" # Optional
export VIEWONLY="false" # Production safety
export MCP_EDITION="viewer" # viewer | editor | admin (default: viewer)# Development Registry
export SCHEMA_REGISTRY_NAME_1="development"
export SCHEMA_REGISTRY_URL_1="http://dev-registry:8081"
export VIEWONLY_1="false"
# Production Registry (with safety)
export SCHEMA_REGISTRY_NAME_2="production"
export SCHEMA_REGISTRY_URL_2="http://prod-registry:8081"
export VIEWONLY_2="true" # Read-only protectionPre-configured examples available in config-examples/:
Configuration | Use Case | File |
Production | Stable Docker deployment | |
Multi-Environment | DEV/STAGING/PROD registries | |
Local Development | Python local execution | |
View-Only Safety | Production with safety |
📖 Complete configuration guide: config-examples/README.md
An edition decides which tools are registered. Tools outside the active edition are not listed and cannot be called, so they cost no context and cannot be invoked by mistake.
Edition | Tools | Use case |
| 39 | Browse, export, compare, compatibility checks |
| 42 | Viewer plus schema registration and evolution |
| 67 | Full surface: configuration, deletion, migration, bulk operations |
export MCP_EDITION="admin"Precedence: MCP_EDITION > SLIM_MODE alias > MCP_EDITION_DEFAULT >
viewer. Setting MCP_EDITION and SLIM_MODE together is a startup error.
Resources are shared by every edition. Read-only access stays available in
viewer through the registry://, schema:// and subject:// resources.
VIEWONLY is enforced at the HTTP session boundary. A read-only registry
rejects PUT, DELETE and non-read POST requests even if a write tool is
exposed by the edition.
The tool surface is generated from tool_catalog.py; the tables in
docs/editions.md are generated from the same catalog.
Edition | Tools | Highlights |
| 39 | Registry info, subjects, schemas, compatibility, export, comparison, statistics, resources |
| +3 |
|
| +26 | Configuration, deletion, contexts, migration, bulk operations, workflows |
The server also exposes 10 prompts, 4 concrete resources and 18 resource templates. See docs/mcp-tools-reference.md for parameter details.
# In Claude Desktop, use natural language:
"Register a user schema with id, name, email fields"
"Check if my updated schema is compatible"
"Export all schemas from staging context"
"List subjects in production context""Compare development and production registries"
"Migrate user-events schema from staging to production"
"Test connections to all registries"
"Show me registry statistics""Clear all schemas from test context"
"Export global schemas for backup"
"Count schemas across all contexts"📖 More examples: examples/ | 📖 Use cases: docs/use-cases.md
# Enable authentication
export ENABLE_AUTH=true
export AUTH_ISSUER_URL="https://your-oauth-provider.com"
export AUTH_AUDIENCE="your-client-id"Supported Providers: Azure AD, Google OAuth, Keycloak, Okta, GitHub
Permission Scopes:
read - View schemas, configurations
write - Register schemas, update configs (includes read)
admin - Delete subjects, full control (includes write + read)
VIEWONLY Mode - Prevent accidental changes in production
URL Validation - SSRF protection with configurable localhost access
Scope-based Authorization - Fine-grained tool-level permissions
Per-Registry Controls - Independent safety settings
📖 Security guide: docs/deployment.md#security
Guide | Description |
Complete tool documentation with examples | |
How to add and remove subject aliases | |
Real-world scenarios and implementation patterns | |
Docker, Kubernetes, cloud platforms, CI/CD | |
VS Code, Claude Code, Cursor setup | |
Ready-to-use Claude Desktop configs | |
Comprehensive testing setup | |
Version history and migration notes | |
Major version features |
Examples - Usage examples and code samples
Scripts - Utility scripts and automation
Helm Charts - Kubernetes deployment
Tests - Test suites and validation
cd tests/
./run_all_tests.sh --quick # Essential tests
./run_all_tests.sh # Complete test suitepython tests/test_docker_mcp.py# From repository root
cd inspector-tests
# Single registry (DEV)
./run-inspector-tests.sh stable
# Multi-registry (DEV + PROD)
./run-inspector-tests.sh multi
# Test a specific Docker tag
DOCKER_VERSION=latest ./run-inspector-tests.sh stable📖 Testing guide: TESTING_SETUP_GUIDE.md
# With docker-compose
docker-compose up -d
# Direct Docker
docker run -d -p 38000:8000 \
-e SCHEMA_REGISTRY_URL=http://registry:8081 \
aywengo/kafka-schema-reg-mcp:stable# Using Helm charts
helm install kafka-schema-mcp ./helm/kafka-schema-reg-mcp📖 Deployment guide: docs/deployment.md
We welcome contributions. Open an issue or pull request on GitHub, and see:
git clone https://github.com/aywengo/kafka-schema-reg-mcp
cd kafka-schema-reg-mcp
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
python kafka_schema_registry_unified_mcp.py♻️ Restore - restore_bundle applies an export document to one registry, dry-run unless told to write, and preserves schema ids. Admin is now 67 tools
🔐 One verifier - Human, client-credentials, and SEP-990 access tokens share one check. Grants are advertised only when the issuer lists them
🚢 Fleet front - Optional process that routes registry= to a shard and holds no registry password
⏱️ Cache hints - Per-resource ttlMs, including schema://{name}/ids/{schema_id} (24 h)
🎛️ Editions - viewer (39 tools), editor (42) and admin (66) from one image; SLIM_MODE is a deprecated alias for editor
📋 MCP 2026-07-28 - Dual-era protocol support with FastMCP native background tasks
🔒 FastMCP 4.1+ - Current MCP SDK 2 stack with the 2026 security fixes
🧭 MRTR workflows - Guided schema evolution, migration, context reorganisation and disaster recovery
🚚 Migration guide - docs/v3-migration-guide.md
🧭 Subject Aliasing - New tools add_subject_alias and delete_subject_alias
🛠️ Fixes - Evolution assistant and import interactive fixes
📦 Enhancements - Continued MCP tool refinements and testing improvements
🗑️ Removed Deprecated Tools - Custom task management tools removed in favor of FastMCP's built-in Docket system
🔒 Security Fixes - Resolved credential exposure in logging
🤖 Interactive Schema Migration - Smart migration with user preference elicitation
💾 Automatic Backups - Pre-migration backup creation
✅ Post-Migration Verification - Comprehensive schema validation
🚀 FastMCP 2.8.0+ Framework - Complete architecture upgrade
📊 MCP 2025-06-18 Compliance - Protocol version at the time of the v2.0 release
🔐 OAuth 2.1 Generic Discovery - Universal provider compatibility
🔗 Resource Linking - HATEOAS navigation in tool responses
📖 Full changelog: CHANGELOG.md | 📖 v2.0.0 features: README-v2.0.0-HIGHLIGHTS.md
🐳 Glama.ai:
🐳 Docker Hub: aywengo/kafka-schema-reg-mcp | 📊 Stats: 39 / 42 / 67 tools by edition, 4 resources + 18 templates, 8 registries, OAuth 2.1, multi-platform
License: MIT | Maintainer: @aywengo | Issues: GitHub Issues
Create a subject alias to an existing subject (not available in SLIM/VIEWONLY).
Inputs: alias (required), existing_subject (required), context, registry.
Check if a schema is compatible with the latest version.
Inputs: subject (required), schema_definition (required), context, registry, +1 more.
Interactive compatibility checking with elicitation for resolution options.
When compatibility issues are found, this tool will elicit resolution preferences from the user.
Inputs: subject (required), schema_definition (required), context, registry, +1 more.
Compare contexts across two registries.
Inputs: source_context (required), source_registry (required), target_registry (required), target_context.
Compare two Schema Registry instances and show differences.
Inputs: source_registry (required), target_registry (required), include_configs, include_contexts.
Count the number of contexts in a registry.
Inputs: registry.
Count the number of schemas in a context or registry.
Inputs: context, registry.
Count the number of versions for a specific schema.
Inputs: subject (required), context, registry.
Get detailed information about a workflow definition
Inputs: workflow_id (required). Returns: result
Export all subjects within a context.
Inputs: context (required), registry, include_config, include_metadata, +1 more.
Export all contexts and schemas from a registry.
Inputs: registry, include_config, include_metadata, include_versions.
Interactive global export with elicitation for export preferences.
When export preferences are not specified, this tool will elicit the required configuration from the user.
Inputs: format, registry, compression, output_format, +4 more.
Export a single schema in the specified format.
Inputs: subject (required), format, context, version, +1 more.
Export all versions of a subject.
Inputs: subject (required), context, registry, include_config, +2 more.
Find schemas that exist in source registry but not in target registry.
Inputs: source_registry (required), target_registry (required), context.
Generate resource URI templates for your specific configuration.
Inputs: context, subject, registry_name.
Get the current default registry.
Inputs: no named parameters.
Get global configuration settings.
NOTE: This tool is maintained for backward compatibility. Consider using the 'registry://{name}/config' resource instead for better performance.
Inputs: context, registry.
Get MCP 2026-07-28 specification compliance status and configuration details.
Returns information about JSON-RPC batching status, protocol version, header validation, and migration guidance.
Inputs: no named parameters.
Get the current mode of the Schema Registry.
NOTE: This tool is maintained for backward compatibility. Consider using the 'registry://mode' resource instead for better performance.
Inputs: context, registry.
Get information about OAuth scopes and permissions.
Inputs: no named parameters.
Get detailed information about a specific registry.
NOTE: This tool is maintained for backward compatibility. Consider using the 'registry://info/{name}' resource instead for better performance.
Inputs: registry.
Get comprehensive statistics about a registry.
Inputs: registry, include_context_details.
Get a specific version of a schema.
NOTE: This tool is maintained for backward compatibility. Consider using the 'schema://{name}/{context}/{subject}' resource instead for better performance.
Inputs: subject (required), context, version, registry.
Get a schema by its globally unique ID.
Inputs: schema_id (required), registry.
Get all versions of a schema for a subject.
NOTE: This tool is maintained for backward compatibility. Consider using the 'schema://{name}/{context}/{subject}/versions' resource instead for better performance.
Inputs: subject (required), context, registry.
Get configuration settings for a specific subject.
NOTE: This tool is maintained for backward compatibility. Consider using the 'subject://{name}/{context}/{subject}/config' resource instead for better performance.
Inputs: subject (required), context, registry.
Get the operational mode for a specific subject.
NOTE: This tool is maintained for backward compatibility. Consider using the 'subject://{name}/{context}/{subject}/mode' resource instead for better performance.
Inputs: subject (required), context, registry.
Get subjects and versions associated with a schema ID.
Inputs: schema_id (required), registry.
Start the Schema Evolution Assistant workflow. This guided workflow helps you safely evolve schemas by analyzing changes, suggesting strategies, and coordinating consumer updates.
Inputs: subject, strategy, current_schema, proposed_schema.
List all available MCP resources and their usage patterns.
Inputs: no named parameters.
List all available schema contexts.
NOTE: This tool is maintained for backward compatibility. Consider using the 'registry://{name}/contexts' resource instead for better performance.
Inputs: registry.
List all configured Schema Registry instances.
NOTE: This tool is maintained for backward compatibility. Consider using the 'registry://names' resource instead for better performance.
Inputs: no named parameters.
List all subjects, optionally filtered by context.
NOTE: This tool is maintained for backward compatibility. Consider using the 'registry://{name}/subjects' resource instead for better performance.
Inputs: context, registry.
List available multi-step workflows
Inputs: no named parameters. Returns: result
Respond to MCP ping requests with pong.
This tool implements the standard MCP ping/pong protocol for server health checking. MCP proxies and clients use this to verify that the server is alive and responding.
Inputs: no named parameters.
Register a new schema version.
Inputs: subject (required), schema_definition (required), context, registry, +1 more.
Interactive schema registration with elicitation for missing field definitions.
When schema_definition is incomplete or missing fields, this tool will elicit the required information from the user interactively.
Inputs: subject (required), context, registry, schema_type, +1 more.
Suggest the appropriate resource to use instead of a removed tool.
Inputs: tool_name (required).
Test connections to all configured registries.
NOTE: This tool is maintained for backward compatibility. Consider using the 'registry://status' resource instead for better performance.
Inputs: no named parameters.
Test OAuth discovery endpoints to ensure proper MCP client compatibility.
Validates:
/.well-known/oauth-authorization-server
/.well-known/oauth-protected-resource
/.well-known/jwks.json
Inputs: server_url.
Test connection to a specific registry.
NOTE: This tool is maintained for backward compatibility. Consider using the 'registry://status/{name}' resource instead for better performance.
Inputs: registry.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 42 tools
Many tools target distinct operations, but there are several overlapping pairs such as register_schema vs register_schema_interactive, check_compatibility vs check_compatibility_interactive, and export_global vs export_global_interactive. Deprecated tools are explicitly noted as replaceable by resources, which creates ambiguity about whether an agent should call the tool or use the resource. The descriptions help, but the set still has notable boundary confusion.
Most tools follow a predictable snake_case verb_noun pattern, including clear prefixes like get_, list_, test_, export_, and count_. The interactive variants use a consistent _interactive suffix. Minor deviations such as ping and guided_schema_evolution keep it from being perfectly uniform.
With 42 tools, the surface is heavy for a Schema Registry server, especially because many tools are explicitly maintained only for backward compatibility and point to resources instead. The large number of near-duplicate interactive/non-interactive variants and deprecated tools makes the set feel bloated rather than well-scoped.
The tools cover many read operations, registration, compatibility checks, exports, counts, statistics, and registry comparisons. However, write-side lifecycle operations such as deleting subjects or schema versions and setting global/subject config or mode appear missing, leaving notable gaps. Agents can perform many workflows but may hit dead ends for administrative mutations.
MEOK MCP Test MCP — golden-file + schema-drift + tool-failure tests for any MCP server. Drop-in
The official MCP Server from Mia-Platform to interact with Mia-Platform Console
MCP server for interacting with the Supabase platform
MCP compatibility and transport verification server.
