Build AI Agents
Without the Magic

GENT is the open-source language for AI agents. Type-safe outputs, radical transparency, and Rust-powered performance.

agent.think()→ action.execute(){ tools }

Install GENT
v0.1.0

Release notes →
brew install gent-lang/tap/gent

macOS and Linux supported. Windows support coming soon.

GENT is the language for production AI agents

Built on web standards with zero-config tooling, unmatched transparency, and a complete built-in toolkit.

RustPowered Runtime
100%Type Safe
0Hidden Transforms

All your favorite tools, built-in

GENT natively supports type-safe outputs, tool definitions, RAG, and multi-agent orchestration with zero configuration.

🔒Type-Safe Outputs

Define structured schemas for LLM outputs. Get compile-time validation and automatic retries when outputs don't match.

Learn more
$ gent run analyzer.gnt
✓ Output validated successfully
{ sentiment: "positive", confidence: 0.95 }
analyzer.gnt
1struct Analysis {
2 sentiment: string
3 confidence: number
4 keywords: string[]
5}
6
7agent Analyzer {
8 model: "gpt-4o"
9 systemPrompt: "Analyze text sentiment."
10
11 output: Analysis
12 outputRetries: 3
13}
14
15let result = Analyzer
16 .userPrompt("I love this product!")
17 .run()
18
19// result.sentiment -> "positive"
20// result.confidence -> 0.95

No hidden magic

Other frameworks inject hidden context, transform your prompts, and add unexpected behavior. GENT gives you complete control.

Other frameworks

$ python agent.py
What gets sent to LLM:
[Hidden system context injected]
[Framework-specific formatting]
[Auto-generated tool schemas]
Your actual prompt (modified)
[Memory context appended]
Hidden transformations affect LLM behavior

GENT

$ gent run agent.gnt
What gets sent to LLM:
System: "You are a helpful assistant."
User: "Hello, help me with this."
What you write is exactly what the LLM sees

High-performance
agent execution

Built in Rust for maximum performance. Execute hundreds of agents concurrently with minimal overhead.

Native Performance

Compiled to native code with zero runtime overhead. No interpreter, no garbage collection pauses.

True Parallelism

No Python GIL or JavaScript event loop limitations. Execute agents on all available CPU cores.

Minimal Memory Footprint

Efficient memory management without a heavy runtime. Run more agents with less resources.

Concurrent by Design

Built from the ground up for concurrent agent execution using Rust's async runtime.

Powered by Rust

Ready to get started with GENT?