RooBot logoAUSSIE ROBOTS●
Explore
Let’s talk ↗
A metallic Turul bird racing along copper and blue data trails, with the headline Up to 20 times faster.
The 20× claim applies to the reported local MCP comparison: Rust client → Turul versus Rust client → FastMCP on an AMD 9900X. It is not a benchmark for every Turul component.

TURUL / BY AUSSIE ROBOTS

One family.
Composable Rust foundations.

Turul brings our open-source agent infrastructure together: tools, agent communication, model access, skills, messaging and token validation. Choose the components your application needs.

Explore the components ↓Browse the source ↗

HOW THE NAMES FIT TOGETHER

COMPANYToast Technology Pty Ltd
ENGINEERING BRANDAussie Robots
OPEN-SOURCE RUST FAMILYTurul

MCP · A2A · LLM · Agent Skills · RPC · JWT

Turul is the family name. Each project keeps its own repository, packages and release cadence. ROS 2 drivers and GPS Trust sit alongside Turul within Aussie Robots.

SIX COMPONENTS / DISTINCT RESPONSIBILITIES

Start with the capability.
Find the right crate.

PROTOCOL IMPLEMENTATIONS

Turul MCP

Connect tools and context

Build Model Context Protocol servers and clients in Rust, with tools, resources and prompts.

Selected crates
  • turul-mcp-server
  • turul-mcp-client
  • turul-mcp-derive
Explore MCP on GitHub ↗
PROTOCOL IMPLEMENTATIONS

Turul A2A

Connect agents

Agent-to-Agent protocol infrastructure for agent discovery, messaging and task interaction.

Selected crates
  • turul-a2a
  • turul-a2a-client
  • turul-a2a-types
Explore A2A on GitHub ↗
MODEL ACCESS

Turul LLM

Connect models

A provider-neutral client trait and adapters for structured model output. Ollama and OpenAI-compatible adapters live in separate crates.

Selected crates
  • turul-llm-core
  • turul-llm-ollama
  • turul-llm-openai
Explore LLM on GitHub ↗
SKILLS TOOLING

Turul Agent Skills

Package agent know-how

Read, validate, create and catalog Agent Skills directories. A library and CLI for the open SKILL.md format, with LLM-backed examples.

Selected crates
  • turul-agent-skills
  • turul-agent-skills-cli
Explore Agent Skills on GitHub ↗
SUPPORTING INFRASTRUCTURE

Turul RPC

Handle structured messages

Transport-agnostic JSON-RPC 2.0 types, codec and async dispatch. Domain errors are mapped to protocol responses at the dispatcher.

Selected crates
  • turul-rpc
  • turul-rpc-core
  • turul-rpc-jsonrpc
  • turul-rpc-server
Explore RPC on GitHub ↗
SUPPORTING INFRASTRUCTURE

Turul JWT

Validate access tokens

Local JWT validation with JWKS caching, shared by the MCP and A2A projects. Authentication is one part of application security.

Selected crates
  • turul-jwt-validator
Explore JWT on GitHub ↗

This is a capability map, not a dependency graph. Check each repository for supported versions, compatibility, licence terms and current release status.

A PRACTICAL COMPOSITION

Skills describe the work.
Frameworks connect the parts.

Give an agent a skill

Turul Agent Skills discovers and validates packaged instructions. Its examples use Turul LLM to select, author and review skills.

Connect its capabilities

Use MCP for tools and context, and A2A for interaction with other agents. Applications choose how these capabilities work together.

Keep boundaries explicit

Typed RPC messages and JWT validation provide supporting infrastructure. Authority, approval and oversight remain application decisions.

The MCP repository also includes developer skills for building servers and clients. Those guidance packages are distinct from the Turul Agent Skills Rust library and CLI.

Explore MCP developer skills ↗

TURUL MCP / PERFORMANCE IN CONTEXT

The bird behind the benchmark.

Named after the Turul bird of Hungarian mythology, the family pairs a shared identity with focused engineering. This artwork accompanies the published Turul MCP benchmark.

Read the benchmark conditions →Reproduce the benchmark ↗

BUILD WITH TURUL

Bring a workflow.
Choose the right foundations.

Discuss an integration ↗