One language. Systems to AI. Native performance.
Python-level readability · Rust-like safety · clang++-class native performance · AI-native workflows
Agam is a compiled programming language and toolchain that unifies systems programming, automation, and AI/numerical computing into one coherent language — without splitting into disconnected sub-languages or relying on foreign library wrappers.
Built on a comprehensive 12+ pillar architecture and executing a rigorous 6-tier development roadmap, Agam provides seamless cross-platform native UI capabilities, hardware-accelerated GPU/NPU performance, and world-class cybersecurity primitives. The project is designed from the ground up to offer the ergonomics of Python with the performance guarantees and safety of Rust and C++.
| Principle | What it means |
|---|---|
| 🐍 Readable | Ergonomic, indentation-significant syntax for everyday code. |
| 🦀 Safe | Rust-like safety, traceable diagnostics, and strong type semantics. |
| 🚀 Fast | Direct LLVM backend optimizing to match or exceed optimized clang++. |
| 🧠 AI-Native | Tensor, autodiff, and numerical workflows are language-native capabilities. |
| 🎯 Unified | One language from basic scripts to scalable systems and scientific computing. |
Agam's compiler architecture scales from quick iteration to robust production deployment.
- Multiple Syntax Modes:
@lang.basefor clean, Python-like readability.@lang.base.dynamicfor scripting and dynamic binding.@lang.advancefor explicit, brace-delimited systems programming.
- Versatile Backends:
- LLVM: The primary AOT product direction for maximum native performance.
- JIT (Cranelift): Fast in-memory execution for testing and local execution loops.
- C Backend: A portable fallback target.
- Hardware Accelerated: Built-in
@gpukernel syntax, natively lowering to NVPTX and integrating GPU fast-math and memory models directly into the semantic layer.
@lang.base
fn main():
let total = 40 + 2
if total == 42:
return 0
return 1
Agam's optimization pipeline, including its unique call-cache specialization layer, delivers highly competitive runtime metrics.
| Target | Runtime | vs. Winner | Memory Footprint |
|---|---|---|---|
| Agam LLVM O3 + Call Cache | 12.5ms | 🏆 Winner | ~3.6 MiB |
| Clang C O3 | 23.5ms | +88% | ~3.6 MiB |
| Clang++ O3 | 22.8ms | +82% | ~3.6 MiB |
| Rust release | 23.8ms | +91% | ~4.0 MiB |
| Go release | 33.8ms | +171% | ~5.6 MiB |
| CPython | 359.2ms | +2778% | ~11.6 MiB |
(Note: Agam's native GPU/NPU @gpu kernel pipeline actively expands our performance advantage in hardware-accelerated parallel workloads).
Agam is executing against a structured, 6-tier roadmap designed to establish it as a world-class production language:
- Tier 0: Foundation Completion — Type system, formal grammar, object model, and module visibility. (Active Priority)
- Tier 1: Developer Experience — Elite LSP error recovery, zero-friction visual toolchain, and unified package management.
- Tier 2: Runtime & Security — World-class cybersecurity (capabilities, taint, formal proofs) and secure supply-chain sandboxing.
- Tier 3: Platform & Ecosystem — Universal FFI (zero-friction interop with C/C++/Python/Java), Omni-platform native rendering, and WASM.
- Tier 4: Performance & Optimization — GPU/NPU completion, hardware introspection, and a 120 FPS hardware-accelerated visual engine.
- Tier 5: AI-Native Differentiation — Autodiff, tensor types, and ML training loop primitives built directly into the language.
- Tier 6: Frontier — Self-hosting and AI-native compiler intelligence.
Agam's ecosystem spans across multiple repositories under the @agam-lang organization to cleanly separate compiler internals from platform layers and community tooling.
agam— The core compiler, CLI (agamc), runtime, and optimization pipelines.std— The Agam standard library (I/O, networking, data structures).registry-index— The central package registry index.sdk-packs— Pre-built SDK and toolchain bundles for multi-platform distribution.rfcs— Language design RFCs and community proposals.
agamlab— A MATLAB-like interactive scientific computing and exploration platform.agam-ml— Machine learning foundations.benchmarks— Comprehensive cross-language benchmark suite covering algorithms, I/O, ML primitives, and GPU offloading.
agam-vscode— Official VS Code extension providing rich syntax, LSP integration, and debugging.agam-intellij— Official IntelliJ/IDEA plugin.agam-cli— Extracted CLI utilities and terminal UI components.playground— Web-based environment to run and share Agam code.
agam-web— Web framework and routing.agam-db— Database drivers and ORM capabilities.agam-http/agam-json/agam-crypto/agam-async
agam-book— The definitive guide to the Agam Programming Language.agam-by-example— Interactive examples to learn Agam syntax and idioms.examples— Curated example projects demonstrating best practices.awesome-agam— A curated list of awesome Agam frameworks, libraries, and software.governance— Organization policies, decision-making frameworks, and steering.
Get started with Agam directly from source:
# Build the compiler from source
cargo build -p agam_driver
# Create a new project
agamc new hello_agam && cd hello_agam
# Run with the best available native or JIT backend
agamc run main.agam --fast
# Explicitly use the Cranelift JIT for rapid iteration
agamc run main.agam --backend jit
# Check local toolchain and LLVM readiness
agamc doctorWe welcome contributions across all repositories! Whether you are interested in compiler internals, standard library development, documentation, or benchmarking, there is a place for you.
Please review our Contributing Guide and Code of Conduct.
Note: Every repository in the Agam ecosystem includes an .agent/ directory containing structured guidance, rule sets, and checklists to support AI-assisted and human collaborative development workflows.
All repositories under the agam-lang organization are dual-licensed under the MIT License and the Apache 2.0 License.
Built with 💜 by the Agam community