Get started
Install Wanaku, run your first governed MCP proxy, and learn the architecture.
Policy, credentials, and audit for every tool call, agent message, and inference request.
Wanaku is a governed action proxy for AI agents, and it works as an MCP gateway. It sits between agents and the systems they act on, and it intercepts tool calls, agent-to-agent messages, and inference traffic. Agents never touch backend systems directly: policy, credentials, data controls, and audit happen in the proxy.
| You want to | Start with | Then read |
|---|---|---|
| Try Wanaku for the first time | Quickstart | Architecture |
| Move from the Java-based Wanaku 0.2 | Migrate from Classic Wanaku | Deploy on Kubernetes and OpenShift |
| Deploy and run Wanaku | Deploy and operate Wanaku | Secure a Wanaku deployment |
| Connect agents and backends | Connect agents and backends | A2A proxy, Inference proxy |
| Decide what agents can do | Govern agent actions | Govern an MCP server |
| Add your own logic | Extend Wanaku | Feature crates, Admin UI plugins |
| Look up a setting or a route | Wanaku reference | Environment variables |
| Contribute code or docs | Contributing to Wanaku | Testing |
Wanaku is a good fit when:
Wanaku is not a good fit yet when:
Read Architecture for the other limits of this release.
These pages describe the Rust-based Wanaku server. Its early-access builds are development builds of the unreleased Wanaku 0.3.0. Releases up to v0.2.0 contain the Java-based Wanaku, which is deprecated. If you use the Java-based Wanaku, read Migrate from Classic Wanaku.