The Year in AI Papers
The year/Together AI

Paper 2602.13692

ThunderAgent: A Simple, Fast and Program-Aware Agentic Inference System

Hao Kang, Ziyang Li, Xinyu Yang, Weili Xu, Yinfang Chen, Junxiong Wang, Beidi Chen, Tushar Krishna, Chenfeng Xu, Simran Arora

Published
Feb 2026
Research lab
Together AI
Citations
13
GitHub
Not linked

01 In brief

Summary

ThunderAgent is a program-aware agentic inference system that improves throughput for multi-turn LLM agent workflows by abstracting them as agentic programs with metadata (ID, context length, tool environments, placement, phase, status).

It introduces a program-aware scheduler with state-aware pausing and dynamic migration across GPU nodes to reduce KV cache thrashing and memory imbalance, and a tool resource manager with hook-based garbage collection and asynchronous environment preparation.

Evaluations on coding, routing, and scientific discovery agents show 1.5-3.6x throughput improvements in serving, 1.8-3.9x in RL rollout, and up to 4.2x disk memory savings compared to state-of-the-art systems like vLLM and Continuum.

The system is open-sourced at https://github.com/ThunderAgent-org/ThunderAgent and has been adopted by frameworks like SkyRL and NVIDIA Dynamo.

02 From the paper

Abstract

Large language models(LLMs) are now used to power complex multi-turn agentic workflows. Existing systems run agentic inference by loosely assembling isolated components: an LLM inference engine (e.g., vLLM) and a tool orchestrator (e.g., Kubernetes). Although agentic workflows involve multiple LLM and tool requests, these systems schedule and allocate resources separately on a per-request basis, without end-to-end knowledge of the workflow. This leads to sub-optimal management of KV cache and tool execution environments. To address the challenges, we propose ThunderAgent, a fast, simple, and program-aware agentic inference system. We first abstract agentic workflows as LLM Programs, enabling a unified view of heterogeneous resources, including KV caches, system states, and external tool assets such as disk memory and network ports. Built upon this abstraction, ThunderAgent introduces a program-aware scheduler and a tool resource manager designed to maximize KV cache hit rates, mitigate memory imbalances, and enable asynchronous environment preparation. Evaluations across coding, routing, and scientific discovery agents demonstrate that ThunderAgent achieves 1.5-3.6x throughput improvements in serving, 1.8-3.9x in RL rollout, and up to 4.2x disk memory savings compared to state-of-the-art inference systems. To facilitate reproducibility and support future development, we open-source the system implementations of the whole ThunderAgent at: https://github.com/Agentic-Kinetics/ThunderAgent.