jikkou

---

Your platform team runs Kafka for everyone else. Fifty app teams, and every new topic is a ticket. Configs drift between staging and production because someone "fixed" something by hand during an incident. Nobody can say who changed what, or why.

Jikkou (jikkou / 実行, execution in Japanese) fixes this the way kubectl fixed Kubernetes: declare your Kafka resources (topics, ACLs, quotas, schemas, connectors) as YAML in Git, review changes in pull requests, and let Jikkou reconcile the desired state against your real clusters. Kafka-first, and platform-ready: the same engine manages Schema Registry, Kafka Connect, Confluent Cloud, Aiven, Amazon MSK, AWS Glue, and Apache Iceberg.

Why Jikkou?

Your problem	Jikkou's answer
"Someone changed a topic by hand and nobody noticed"	Stateless reconciliation. No state file; jikkou diff compares Git against the actual cluster, so drift is always visible
"300 topics that differ by one value"	Jinja templating. One definition, per-environment values files
"Same config must land on five clusters"	Multi-cluster apply. Provider groups target a whole fleet in one command
"Every change needs review and an audit trail"	GitOps-native. YAML in Git, diff in the pull request, apply in CI
"We can't let a typo reach production"	Validations & dry-run. Platform rules (naming, min ISR, partition limits) enforced before anything is applied
"We also run Schema Registry, Connect, Iceberg…"	One agnostic engine. Pluggable providers, one resource model across your whole data platform

How Jikkou Compares

• Jikkou vs Terraform: Terraform provisions clusters; Jikkou manages what lives inside them, without a state file to drift.
• Jikkou vs Strimzi Topic Operator: on Strimzi, keep the Topic Operator for topics; Jikkou covers everything it can't reach.

Quick Start

Install

# Via SDKMan (recommended)
sdk install jikkou

# Or via Docker
docker pull streamthoughts/jikkou

See the full installation guide for native binaries, Homebrew, and more.

Define a Kafka topic

# kafka-topics.yaml
apiVersion: 'kafka.jikkou.io/v1beta2'
kind: 'KafkaTopic'
metadata:
  name: 'my-topic'
spec:
  partitions: 12
  replicas: 3
  configs:
    min.insync.replicas: 2

Preview, then apply

# Preview the change against the real cluster, not a state file
jikkou diff --files ./kafka-topics.yaml

# Apply it
jikkou apply --files ./kafka-topics.yaml

That's it. Jikkou computes the diff and applies only the necessary changes:

TASK [CREATE] Create a new topic my-topic (partitions=12, replicas=3) - CHANGED
EXECUTION in 2s 661ms
ok: 0, created: 1, altered: 0, deleted: 0, failed: 0

One Model, Every Platform

The same declarative YAML model covers your entire streaming platform. And because the engine is platform-agnostic, it extends beyond Kafka: Apache Iceberg tables, views, and namespaces are managed with the exact same workflow.

Apache Kafka	Schema Registry	Kafka Connect	Apache Iceberg	Cloud Providers
Topics & Configs	Avro Schemas	Connectors	Tables	Aiven (ACLs, Quotas)
ACLs	JSON Schemas		Views	Confluent Cloud (RBAC)
Quotas	Protobuf Schemas		Namespaces	AWS Glue Schemas
Consumer Groups
Brokers & Users
KTable Records

How It Works

Jikkou follows a simple reconciliation loop:

1. Read your resource definitions from YAML files (with Jinja templating support)
2. Compute the differences between desired state and actual cluster state
3. Apply only the minimal set of changes needed
4. Report what was created, updated, or deleted

Deployment Modes

Mode	Description
CLI	Run as a command-line tool — perfect for local development and CI/CD pipelines
API Server	Run as a REST API server — ideal for platform teams and automation
Docker	Available as container images on Docker Hub
Native Binary	GraalVM-compiled native executables for instant startup

Use Jikkou with AI Agents

Jikkou ships official Agent Skills so coding agents can manage your Kafka resources through the CLI — with validation and dry-runs built into the workflow.

Claude Code:

/plugin marketplace add streamthoughts/jikkou
/plugin install jikkou

Other agents: copy skills/managing-kafka-resources into your agent's skills directory.

The skill never applies changes without showing a diff or dry-run first, and asks before installing anything.

Documentation

Full documentation is available at jikkou.io.

• Getting Started
• Concepts & Architecture
• Providers Reference
• Tutorials

Developers

For build instructions, development setup, and contribution guidelines, see:

• CONTRIBUTING.md — How to contribute, coding guidelines, commit conventions
• AGENTS.md — Detailed development guidelines, build commands, and architecture

Quick Build

# Build and run all tests
./mvnw clean verify

# Build without tests
./mvnw clean verify -DskipTests

# Apply code formatting
./mvnw spotless:apply

Requirements: Java 25, Docker (for integration tests), GraalVM (for native builds)

Contributors

Jikkou is built by its community. Thank you to everyone who has contributed!

Want to see your name here? Check out the contribution guide and open issues.

Support the Project

If you find Jikkou useful, please consider:

• Using Jikkou in production? Add your organization to ADOPTERS.md or tell us in GitHub Discussions: real-world usage reports help the project more than anything else
• Giving it a star on GitHub to help others discover it
• Joining the Slack community to ask questions and share feedback
• Contributing code, documentation, or bug reports

License

Licensed under the Apache License, Version 2.0.

---