Command line
The kurkik command, its terminal UI and its MCP server. It ships with its own JVM, so there is nothing to install first.
$ kurkik doctorKurkik is the messaging-infrastructure side of a self-maintaining API. It reads your on-premise or cloud brokers, writes what it finds as AsyncAPI, and catches drift between your documents and what actually runs - then proposes the fix.

# what prod holds$ kurkik topology prod --consumers# staging, written down as AsyncAPI$ kurkik export stagingwrote staging-kafka.yaml# and what prod does not agree with$ kurkik diff staging-kafka.yaml prod; echo exit $?exit 2 # drift$ 
$ kurkik# preview a Kafka, a Pulsar, an SNS and an SQS connection# every topic setting and where it comes from, partitions, schema, who reads it# walk a diff of staging against prod, as a tree# add a connection in a form, export with a choice$ $ claude mcp add kurkik -- kurkik mcp> Orders stopped reaching the billing service on prod. Compare prod with staging and tell me what drifted.asyncapi_connections asyncapi_topology prod asyncapi_compare staging prodThe evidence: the topic and its consumers, side by side, read-only.$ One AsyncAPI document for every broker you run
A self-maintaining API finds drift as soon as it appears and gives you the tools to find the reason, prepare a patch and apply it. Kurkik covers the messaging systems; the AsyncAPI plugin for JetBrains IDEs covers your source code.
Read-only by default. The only writes are syncs, and those show you the plan first.

$ kurkik topology prod --consumers$ kurkik topology queues-prod# or all of it, browsable$ kurkik$ 
$ kurkik export staging --overwritewrote staging-kafka.yaml$ $ kurkik diff staging prod --consumers --kinds changed,missing,unmanaged$ kurkik diff shop-staging shop-prod --consumers$ kurkik diff queues-staging queues-prod$ $ kurkik kafka sync staging-kafka.yaml prod --dry-run# the plan, and the broker’s verdict on it; nothing changes$ kurkik kafka sync staging-kafka.yaml prod --overwrite$ kurkik kafka sync history prod$ kurkik kafka sync revert <snapshot>$ $ claude mcp add kurkik -- kurkik mcp> Compare staging and prod and tell me what drifted.$ # in your pipeline$ kurkik diff asyncapi.yaml stagingexit 2 # drift: the build stops here$ One engine underneath, so the same connection gives the same answer everywhere.
The kurkik command, its terminal UI and its MCP server. It ships with its own JVM, so there is nothing to install first.
$ kurkik doctorKurkik powers the AsyncAPI plugin: connections, topology and export without leaving your IDE. Connections move to the CLI in one command.
$ kurkik conn import --from-ideaClaude Code, Codex or any agent that speaks MCP. The tools only read, and secrets are redacted before an agent sees them.
$ claude mcp add kurkik -- kurkik mcp$ curl -fsSL https://raw.githubusercontent.com/Pakisan/kurkik-app/master/install.sh | shApple silicon and Intel. Kurkik brings its own JVM, so nothing else needs installing. Then run kurkik doctor.
The script installs into ~/.kurkik, checks the archive's checksum and, when gpg is installed, its release signature, and adds kurkik to your PATH. Read the script before you run it, or pin a version with | sh -s -- --version=1.0.0-eap.1.
$ curl -fsSL https://raw.githubusercontent.com/Pakisan/kurkik-app/master/install.sh | shx64 and arm64, with glibc (Alpine is not supported). Kurkik brings its own JVM, so nothing else needs installing. Then run kurkik doctor.
The script installs into ~/.kurkik, checks the archive's checksum and, when gpg is installed, its release signature, and adds kurkik to your PATH. Read the script before you run it, or pin a version with | sh -s -- --version=1.0.0-eap.1.
x64. Unzip it anywhere, add its bin folder to PATH, and run kurkik from a new terminal. Scoop and winget come later. Kurkik brings its own JVM, so nothing else needs installing. Then run kurkik doctor.
Other platforms and earlier versions: all releases, with checksums and signatures.
Installing Kurkik means you accept the Terms of Use.
No need to start on your own clusters. The demo runs a staging and a prod of Kafka, Pulsar, SNS and SQS in Docker on your machine, already drifted apart, so every command has something to find. Kurkik only reads there, and your own kurkik settings are left alone.

$ kurkik doctor$ kurkik conn lseight connections, all untrusted$ kurkik trust$ kurkik conn lsall ok$ kurkik topology prod --consumers$ kurkik$ git clone https://github.com/Pakisan/kurkik-app && cd kurkik-app docker compose -f demo/docker-compose.yml up -d --wait# the systems; the first start pulls about 3 GB demo/seed.sh# fills them with an "orders" platform source demo/env.sh && cd demo# kurkik’s files stay under demo/.state kurkik trust# what the project’s connections contact; answer y kurkik export staging && kurkik diff staging-kafka.yaml prod# exit 2: driftNeeds kurkik, Docker with Compose v2 and about 5 GB of memory for Docker - or start one system at a time. macOS or Linux; WSL on Windows.
No surprises after you sign up: every integration says where it stands.
Through the AsyncAPI plugin today; deeper integration is planned.
NATS, MQTT, Solace or something in-house - tell us what you run. Custom integrations are on the table.
Let's talkYes. Kurkik writes the first document for you from what is already running. To read, validate and visualise AsyncAPI documents, see the tools at asyncapi.pavelon.dev.
Apache Kafka, Apache Pulsar, Amazon SNS and Amazon SQS. RabbitMQ is next. Schemas are read from Confluent-compatible registries: Confluent, Karapace, Redpanda and Apicurio.
Tell us what you run through the form below. Custom integrations are on the table.
Through its MCP server or a skill. The agent forms a first hypothesis, collects topology and compares environments with Kurkik’s read-only tools, then uses that evidence to refine or reject the hypothesis.
Kurkik only reads - except for kafka sync and pulsar sync. Those show a dry-run plan, ask before they write, take a snapshot first and can be reverted. Credentials are never stored in plain text, and the MCP tools only read.
Kurkik is vendor-neutral: it is not tied to one broker vendor or platform. It speaks one recognisable format, AsyncAPI, so the documents it produces can also configure API gateways or feed docs in EventCatalog.
Yes: the first early-access (EAP) release is out, and Kurkik is free during early access. Install it with one command. Each EAP release works until a date fixed when it is built, and a newer one comes out before then. Not ready to point it at your own brokers? Start in the demo sandbox. Leave your email to hear when a new release is out.
Kurkik is in early access, and new EAP and release-candidate builds follow. Leave your email to hear when they are out, to be first in line for the next one, or to ask for the messaging system you need. No newsletter, and nothing shared or sold.