Danube Messaging

Try Danube

Download the latest danube binary from the releases page or pull the container image.

Standalone

Start a broker and test with danube ctl:

# Start the broker (Client gRPC, Schema Registry, and Admin gRPC share port 6650)
danube up --mode standalone --data-dir ./danube-data

# Produce messages
danube ctl produce -s http://127.0.0.1:6650 -t /default/demo -c 10 -m "Hello, Danube!"

# Consume messages
danube ctl consume -s http://127.0.0.1:6650 -t /default/demo --subscription my_sub

Or run a single broker in a container:

docker run -p 6650:6650 -p 9040:9040 \
  ghcr.io/danube-messaging/danube:latest \
  up --mode standalone --data-dir /data --broker-addr 0.0.0.0:6650

Cluster

Start a zero-YAML 3-node Raft consensus cluster directly with CLI flags:

danube up --mode cluster \
  --broker-addr 0.0.0.0:6650 --raft-addr 0.0.0.0:7650 \
  --data-dir ./danube-data \
  --seed-nodes "node1:7650,node2:7650,node3:7650"

Or launch a local 3-broker cluster using Docker Compose:

git clone https://github.com/danube-messaging/danube.git
cd danube/docker
docker compose up -d --build

Check cluster health (defaults to http://127.0.0.1:6650):

danube ctl brokers list
danube ctl cluster status

Edge (IoT)

Zero-YAML MQTT gateway at the edge (# -> /edge1/ingest out of the box). Validates payloads, buffers into a local WAL during outages, and replicates to your central cluster over gRPC:

danube up --mode edge \
  --edge-name edge1 \
  --cloud-endpoint http://cluster:6650 \
  --mqtt-addr 0.0.0.0:1883 \
  --data-dir ./edge-data

---

Why Danube?

One Binary, Zero YAML

Run as Standalone, Cluster, Edge, or Admin/MCP all from the single danube binary (danube up, danube ctl, danube mcp) with single-port multiplexing on 6650 and no YAML files required. Learn more →

Intent-Driven Pub/Sub & Queuing

Create topics with declarative workload intents (realtime, event_stream, task_queue, audit_log) or explicit Reliable / NonReliable dispatch, WAL sync, retries, and dead-letter queues. Learn more →

Resilient Raft Clusters

Embedded Raft consensus stores cluster policies dynamically in /cluster/config/*, automates leader election and topic reconciliation, and balances hotspots with the load manager. Learn more →

Cloud-Native Tiered Storage

Local WAL for fast writes with background export to S3, GCS, Azure Blob, or shared filesystems, plus zero-copy topic branching and time-travel seeks. Learn more →

Schema Registry

Built-in schema management with JSON Schema, Avro, and Protobuf, live schema inference (topics infer-schema), and backward, forward, and full compatibility enforcement. Learn more →

AI-Native Control Plane

Operate and debug clusters via danube mcp: 71 tools, 4 guided workflows (including explain_consumer_stall), declarative apply_desired_state with dry_run guardrails, and built-in 1-hour telemetry and log ring buffers—no Prometheus required. Learn more →

---

Explore

Danube for AI Agents Workload intents, agent runtime primitives (extend_lease, report_progress, cancel_task_tree, request_reply), zero-copy branching, and MCP operations.

Configuration Workflow Clean 4-layer configuration: local node flags, Raft cluster state, GitOps apply, intent-driven topics, and break-glass recovery.

Client Libraries Rust, Go, Java, and Python SDKs with producer, consumer, agentic primitives, and schema registry examples.

Security Zero-trust auto-TLS, manual TLS/mTLS encryption, JWT authentication, and fine-grained RBAC with default-deny semantics.

Edge & IoT Zero-YAML MQTT gateway with schema validation at the edge and WAL-buffered replication to the central cluster.

Lakehouse Export Continuously export streaming data to Apache Iceberg tables, queryable by Spark, Trino, DuckDB, and any Iceberg-compatible engine.

---

Client Libraries

Rust · crates.io · examples   |   Go · pkg.go.dev · examples   |   Java · Maven Central · examples   |   Python · PyPI · examples