Before starting: If any tool call fails with "Missing required environment variables", tell the user:
"Set your Confluent Cloud API credentials — generate at https://confluent.cloud/settings/api-keys, then export CONFLUENT_CLOUD_API_KEY=... CONFLUENT_CLOUD_API_SECRET=... and restart Claude Code."
Help me set up a complete Confluent Cloud streaming application from scratch.
This is an end-to-end workflow. Use AskUserQuestion interactive forms throughout to collect configuration. Walk me through each step, confirming before proceeding to the next.
Step 1: Resolve or Create Environment
Read .confluent-context.json in the project root. If it has an environment_id, use AskUserQuestion to ask:
Use existing environment? (header: "Environment")
- Option 1: "Use {name} ({id})" (description: "Continue with the active environment")
- Option 2: "Create new" (description: "Set up a fresh environment")
If creating new, use AskUserQuestion in a single call:
Question 1 — Environment name (header: "Name")
- Options: "development" (description: "Good for testing and experimentation"), "staging" (description: "Pre-production environment"), "production" (description: "Production workloads")
- The user can also type a custom name
Question 2 — Stream Governance package (header: "Governance")
- Option 1: "Essentials (Default)" (description: "Free to start, $0.002/schema/hr after 100 free schemas. Includes Schema Registry.")
- Option 2: "Advanced (Recommended)" (description: "$1/hr, $0/schema/hr with up to 20k schemas. Enterprise-ready governance.")
Call create_environment with:
display_name: the chosen namestream_governance_package: "ESSENTIALS" if Essentials selected, "ADVANCED" if Advanced selected
Display the Environment ID, name, and governance package. Save to .confluent-context.json.
Step 2: Create Kafka Cluster
Step 2a: Gather cluster configuration (part 1 — name, type, cloud)
Use the AskUserQuestion tool to collect name, type, and cloud provider in a single call with 3 questions:
Question 1 — Cluster name (header: "Name")
- Options: "my-kafka-cluster" (description: "Default cluster name"), "dev-cluster" (description: "Development cluster"), "prod-cluster" (description: "Production cluster")
- The user can also type a custom name
Question 2 — Cluster type (header: "Type")
- Option 1: "Basic" (description: "First eCKU free, then $0.14/eCKU-hr. 99.5% SLA. Great for dev/test.")
- Option 2: "Standard" (description: "$0.75/eCKU-hr. 99.9-99.99% SLA. For production workloads.")
- Option 3: "Enterprise (Coming Soon)" (description: "$1.75/eCKU-hr. Private networking, self-managed encryption.")
- Option 4: "Dedicated (Coming Soon)" (description: "$2.66/CKU-hr. Fully customizable networking and throughput.")
Question 3 — Cloud provider (header: "Cloud")
- Option 1: "AWS" (description: "Amazon Web Services")
- Option 2: "GCP" (description: "Google Cloud Platform")
- Option 3: "Azure" (description: "Microsoft Azure")
Step 2b: Gather cluster configuration (part 2 — region)
After receiving the cloud provider selection, use AskUserQuestion with a single question showing regions specific to the chosen provider. The user can always type a custom region.
If AWS selected (header: "Region"):
- Option 1: "us-east-1" (description: "N. Virginia (default)")
- Option 2: "us-west-2" (description: "Oregon")
- Option 3: "eu-west-1" (description: "Ireland")
- Option 4: "eu-central-1" (description: "Frankfurt")
If GCP selected (header: "Region"):
- Option 1: "us-east1" (description: "South Carolina (default)")
- Option 2: "us-west1" (description: "Oregon")
- Option 3: "europe-west1" (description: "Belgium")
- Option 4: "europe-west2" (description: "London")
If Azure selected (header: "Region"):
- Option 1: "eastus" (description: "East US (default)")
- Option 2: "westus2" (description: "West US 2")
- Option 3: "westeurope" (description: "Netherlands")
- Option 4: "southeastasia" (description: "Singapore")
Step 2c: Handle special selections
If Enterprise or Dedicated selected: Tell the user: "Enterprise and Dedicated cluster types are coming soon to the Claude Code plugin. For now, you can create these through the Confluent Cloud Console." Then re-ask with only Basic and Standard as options.
If Standard selected: Use AskUserQuestion for one more question:
Uptime SLA (header: "SLA")
- Option 1: "99.9% (Recommended)" (description: "Single zone. Good for development or staging.")
- Option 2: "99.99%" (description: "Multi zone. Recommended for production. Minimum 2 eCKUs.")
Step 2d: Confirm and create
Show a summary in the Confluent Cloud console style:
Summary
─────────────────────────────
Cluster type: Basic
Provider: Amazon Web Services
Region: N. Virginia (us-east-1)
Uptime SLA: 99.5%, Single zone
Networking: Internet
Minimum eCKUs: 1
Ask for confirmation, then call create_cluster with:
- The chosen display_name
- The chosen environment_id
- The chosen cloud provider
- The chosen region
- cluster_type: "BASIC" or "STANDARD"
- availability: "SINGLE_ZONE" (or "MULTI_ZONE" if Standard with 99.99% SLA)
Step 2e: Wait for provisioning
The cluster will start in PROVISIONING status. Poll using get_cluster every 30 seconds until status.phase is PROVISIONED. Show progress updates. Timeout after 15 minutes.
Step 3: Create API Key
Once cluster is provisioned, automatically call create_api_key. Credentials are automatically saved to .confluent-context.json for use by topic commands. Display:
- API Key ID
- API Secret
- Note that credentials have been saved to
.confluent-context.json
- Strongly remind to save the secret — it's only shown once! (The secret is stored in the local context file, but save it separately for use outside Claude Code.)
Step 4: Create Topics
Step 4a: Resolve cluster API key
First, check .confluent-context.json for a stored API key for this cluster (under api_keys[cluster_id]). If found, use it automatically and tell me: "Using stored API key {api_key} for cluster {cluster_id}."
If no stored key is found, call list_api_keys with the cluster's resource_id to check for existing cluster-scoped API keys.
- If keys exist, offer to create a new one with
create_api_key (credentials are saved automatically).
- If no keys exist, offer to create one with
create_api_key. Credentials are saved automatically to context.
Step 4b: Gather topic configuration
Use AskUserQuestion in a single call:
Question 1 — Topic name (header: "Topic")
- Option 1: "events" (description: "General event stream")
- Option 2: "commands" (description: "Command/request stream")
- Option 3: "notifications" (description: "Notification stream")
- The user can also type custom topic names (e.g.,
orders.created, user.events)
Question 2 — Partitions (header: "Partitions")
- Option 1: "6 (Recommended)" (description: "Good default for most development and production use cases")
- Option 2: "1" (description: "Single partition, useful for strict ordering")
- Option 3: "12" (description: "Higher parallelism for high-throughput topics")
Question 3 — Retention (header: "Retention")
- Option 1: "7 days (Recommended)" (description: "Default retention, good for most use cases")
- Option 2: "1 day" (description: "Short retention for transient data")
- Option 3: "30 days" (description: "Longer retention for replay/audit")
- Option 4: "Infinite" (description: "Never delete messages. Set retention to -1.")
Step 4c: Create the topic(s)
Call create_topic with the cluster ID, environment ID, topic name, partitions count, and retention_ms:
- 7 days = 604800000
- 1 day = 86400000
- 30 days = 2592000000
- Infinite = -1
Cluster API credentials are auto-resolved from context — no need to pass them explicitly.
After creating, ask if the user wants to create additional topics. If yes, repeat Step 4b.
Step 5: Scaffold Application (Optional)
Use AskUserQuestion:
Scaffold a starter app? (header: "Scaffold")
- Option 1: "Node.js / TypeScript" (description: "Producer + consumer using kafkajs")
- Option 2: "Python" (description: "Producer + consumer using confluent-kafka")
- Option 3: "Java" (description: "Producer + consumer using Kafka client")
- Option 4: "Skip" (description: "No application scaffolding needed")
If not skipped, create producer + consumer files with:
- Connection config pointing to the new cluster
- Placeholder for API key/secret (reference environment variables)
- Basic error handling and comments
Step 6: Summary
Show a final summary:
Confluent Cloud Streaming App — Setup Complete
Environment: <name> (<env-id>)
Cluster: <name> (<cluster-id>)
Cloud/Region: <cloud> / <region>
Type: <type> (<availability>)
Endpoints:
Bootstrap: <bootstrap-url>
REST: <rest-endpoint>
API Key: <key-id> (secret was shown above — make sure you saved it!)
Topics:
- <topic-1> (<partitions> partitions, <retention> retention)
- <topic-2> (<partitions> partitions, <retention> retention)
Application files: (if scaffolded)
- producer.<ext>
- consumer.<ext>
Suggest what to do next:
- "Run the producer to send test messages"
- "Open the Confluent Cloud console to monitor your cluster"
- "Create additional topics with
/topics-create"