[Paper] Agint: Agentic Graph Compilation for Software Engineering Agents

Source: arXiv

Abstract

LLM-based coding agents are increasingly common but still face challenges in context management, latency, reliability, reproducibility, and scalability. We present Agint, an agentic graph compiler, interpreter, and runtime that incrementally and hierarchically converts natural‑language instructions into typed, effect‑aware code DAGs. Agint introduces explicit type floors (text → data → spec → code) grounded in semantic graph transformations and a hybrid LLM‑and‑function‑based JIT runtime. This enables dynamic graph refinement, reproducible and optimizable execution, speculative evaluation, and interoperability with existing developer tools.

Agint’s typed graph bindings improve reliability and allow concurrent composition of codebases by construction, supporting accelerated development with smaller and faster models, lower latency, efficient context utilization, and higher throughput. Hierarchical compilation allows scalable graph edits, while the graph structure supports reproducibility and efficient parallel generation.

Agint provides a composable Unix‑style toolchain:

• dagify – DAG compiler
• dagent – hybrid JIT runtime
• schemagin – schema generator
• datagin – data transformer

These components enable realtime, low‑latency code and data‑flow creation. Human developers and coding agents refine graphs through the Agint CLI, while non‑technical users use Agint Flow GUI for visual editing, conversational refinement, and debugging, promoting prototype agentic workflows to production code. This continuous co‑creation model allows teams to prototype quickly, refine seamlessly, and deploy reliably, bridging natural language, compiler methods, and developer tooling to enable a new generation of composable, team‑centric coding agents at scale.

Comments

• 18 pages, 5 figures
• Submitted to NeurIPS 2025: Deep Learning for Code in the Agentic Era

Subjects

• Software Engineering (cs.SE)
• Machine Learning (cs.LG)

Citation

@article{chivukula2025agint,
  title   = {Agint: Agentic Graph Compilation for Software Engineering Agents},
  author  = {Abhiram Chivukula},
  journal = {arXiv preprint arXiv:2511.19635},
  year    = {2025},
  url     = {https://arxiv.org/abs/2511.19635}
}

• arXiv: 2511.19635
• DOI: 10.48550/arXiv.2511.19635

Submission History

• v1 – Mon, 24 Nov 2025 19:10:47 UTC (1,554 KB) – submitted by Abhiram Chivukula.