Skip to main content

Apache Iggy 0.5.0 Release

· 3 min read
Piotr Gankiewicz
Apache Iggy founder

Release 0.5.0​

It’s been four months since our last release 0.4.300. Over 150 commits, a dozen new contributors, and countless hours later, we’re proud to announce Apache Iggy 0.5.0!

This release brings major improvements across the board — core server enhancements, updated SDKs, new tooling, data management interfaces, and more.

Iggy Server​

The most significant change on the server side is the introduction of zero-copy (de)serialization. This boosts performance (increased throughput and lower tail latency) while reducing memory usage by avoiding unnecessary allocations and copies.

You can read the full deep dive, but in short — this is a breaking change. It required SDK refactoring, and is supported only in the most recent client library releases.

The new schema is already supported in the Iggy CLI and Iggy Web UI. A simple update is all you need.

Docker images for the server and related components are available on Docker Hub.

SDKs​

Alongside migrating repositories to the monorepo, we updated all SDKs to support the zero-copy schema and fill in missing features. Our supported client libraries now include:

C++ and Elixir clients are in progress under the foreign directory.

New in Rust SDK: Support for a background producer, which buffers messages in the background according to configured linger time and other settings. Examples are available here.

Benchmarks​

We’ve expanded our benchmarking toolkit:

  • Iggy Bench CLI (iggy-bench) — run local benchmarks with various configurations and profiles.
  • JSON output and chart rendering for visual performance analysis.
  • Open benchmark results published at benchmarks.iggy.apache.org, powered by the open source benchmarking suite.

As always, you can install the regular CLI with: cargo install iggy-cli.

Connectors​

We’ve built a new connectors runtime for easily creating Rust-based plugins that integrate Iggy with external data sources—similar to Kafka Connect or Redpanda Connect.

You can:

  • Ingest data into Iggy from external systems (e.g., Postgres CDC).
  • Export data from Iggy to external destinations (e.g., indexing engines).
  • Apply optional transformations and schema mappings along the way.

Learn more in this post and in the documentation.

MCP Server​

Apache Iggy now includes support for the Model Context Protocol (MCP)—an open standard for providing context to LLMs.

You can start the MCP server with cargo run --bin iggy-mcp or use the provided Docker image.

Configuration, credentials, and permissions are detailed in the documentation.

Versioning​

Releases follow semver (e.g., 0.5.0) or the latest Docker tag (apache/iggy:latest).

We also publish edge/dev/nightly builds (e.g., 0.5.0-edge.1 or apache/iggy:edge), which contain the newest changes but are not recommended for production.

What’s Next​

Work is underway on a complete rewrite of the server’s core architecture using the io_uring runtime with a thread-per-core, shared-nothing design. This will further improve performance, reduce tail latencies, and lower resource usage.

Other upcoming efforts:

  • Expanding SDKs across more languages and catching up with the latest features.
  • Synchronous SDK clients and sans-io design.
  • Benchmarking orchestrator
  • Enhancing the Connectors Runtime and Web UI.
  • Extending AI features around the MCP server.
  • Exploring Agentic AI use cases for streaming-based agent coordination.

Thanks to our amazing community and contributors for making Apache Iggy better with every release.

Join us on Discord and help shape the future of the project!