cs.LGAug 20, 2026

AgentDecarbonizer: Carbon-Aware Execution for AI Agents

Authors: Leyi Yan, Shuangning Li, Sihang Liu

Organizations: University of Waterloo, Canada · University of Chicago, USA

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

AI agents extend large language models from single prompt-response interactions to long-running, goaldirected workflows that issue many model calls, invoke tools, and interact with external environments. These workflows enable tasks such as software repair, data analysis, and experiment management, but their repeated model invocations can incur substantial carbon emissions. This paper characterizes the carbon emissions of OpenClaw agent workloads using WildClawBench, and shows that emissions depend on token consumption, context cache reuse, and the carbon intensity of the grid. Our characterization identifies deadline flexibility as an opportunity for carbon-aware execution: agent tasks can wait for lower-carbon-intensity periods or shift to lower-carbon grids. However, doing so requires handling uncertain execution time for temporal shifting and cached context recomputation during spatial shifting. We present AgentDecarbonizer, a carbon optimizer for AI agents that runs alongside OpenClaw. Given a task prompt and user-specified deadline, AgentDecarbonizer conservatively estimates task duration and selects deadline-feasible execution schedules, while accounting for cache recomputation overhead during spatial shifting. Evaluated on WildClawBench workloads with 60 agent tasks across four grids, AgentDecarbonizer reduces carbon emissions by up to 57.9 % compared with a carbon-agnostic baseline and by up to 37.5 % compared with a baseline that selects the carbon-optimal grid at task start time.

Figures & tables

Figure 1. Example of an autonomous AI agent workflow on the OpenClaw platform [ 63 ] .
Figure 2
Category# TasksExample tasks
Productivity Flow10Paper digest, PDF analysis
Code Intelligence12Website development, code debugging
Social Interaction6Chat extraction, calendar scheduling
Search & Retrieval11Data collection, fuzzy search
Creative Synthesis11Slides generation, posters generation
Safety Alignment10Skill injection filtering, password leak detection
Table 2. WildClawBench tasks and categories [ 13 ] .
Figure 4
Figure 5
Figure 7. AgentDecarbonizer overview.
Figure 8. Execution time estimator workflow.
SymbolDescription
R\mathcal{R}Candidate execution regions.
H\mathcal{H}Deadline-feasible candidate intervals.
HHNumber of deadline-feasible candidate intervals.
KKNumber of ordered execution intervals induced by T~\widetilde{T} .
rkr_{k}Region assigned to execution interval kk .
tkt_{k}Candidate interval assigned to execution interval kk .
Table 3. Notation in the AgentDecarbonizer placement objective in Equation 6 .
Figure 9. (a) Overall carbon savings, (b) grid usage, and (c) carbon intensity distribution. Model is GPT-5.4.
Figure 10. Example of agent execution optimized by AgentDecarbonizer. Start time Feb 4, 2026 at 16:00 UTC.
Figure 11. (a) Execution time, (b) token usage, and (c) carbon emissions of Gemini 3.1 Pro vs GPT-5.4.
Figure 12. Emissions of single-grid only scenarios compared to AgentDecarbonizer. The model API is GPT-5.4.
Figure 13
Figure 14
Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

CategoryTypical steps/rangeReference
Single-context text transformation1IFEval [ 97 ]
Structured strict-format output4ToolLLM [ 68 ]
External retrieval/browser/API workflow7.3Mind2Web [ 12 ]
Multi-turn communication/coordination8.93–15.39MultiWOZ [ 10 ]
Repository/code/debug workflow14.71SWE-agent [ 87 ]
Multimodal visual/computer-use workflow15OSWorld [ 82 ]
Appendix
Table 4. Phase categories and number of steps per phase.
More in this topic

Carbon-Aware Computing

See all 20 papers

Explore similar work

CardsList
Sep 11, 2026

Carbon-Aware Routing for Function Calling in Edge-Cloud LLM Systems

Large Language Models (LLMs) with function-calling capabilities are becoming critical for modern agentic AI systems. Nevertheless, current deployments typically route inferences to powerful cloud-based models, incurring significant energy use and carbon emissions. We address this sustainability challenge with a carbon-aware routing framework that distributes function-calling queries across a three-tier edge-cloud architecture, combining edge and cloud LLMs on heterogeneous hardware. At its core, a lightweight k-NN predictor operating in a unified semantic-lexical embedding space estimates query-specific accuracy, delay, and power consumption on each edge tier. These predictions are then combined with real-time grid carbon intensity to route every query to the lowest-emission tier capable of executing it successfully. Evaluated on state-of-the-art function-calling benchmarks and LLM families, our framework matches cloud-level accuracy while reducing operational carbon emissions by 4×4\times on average.

Aikaterini Maria Panteleaki, Varatheepan Paramanayakam, Spyros Tragoudas +1

School of Electrical, Computer and Biomedical Engineering Southern Illinois University Carbondale

LLM RoutingCarbon-Aware ComputingEdge-Cloud Computing
AskPDF
May 12, 2026

GAR: Carbon-Aware Routing for LLM Inference via Constrained Optimization

The growing deployment of large language models (LLMs) makes per-request routing essential for balancing response quality and computational cost across heterogeneous model pools. Current routing methods rarely consider sustainable energy use and CO2 emissions as optimization objectives, despite grid carbon intensity varying by time and region, and models differing significantly in energy consumption. To address this gap, we introduce Green-Aware Routing (GAR), a constrained multi-objective optimization framework that minimizes per-request CO2 emissions subject to explicit accuracy floors and p95-latency service-level objectives (SLOs). GAR employs adaptive constraint optimization through per-dataset floor tuning and incorporates lightweight estimators for correctness, tail latency, and carbon emissions, enabling real-time routing decisions without additional inference passes. We present GAR-PD, a practical online primal-dual routing algorithm for rolling carbon budgets, alongside heuristic variants that achieve high feasibility coverage while limiting accuracy degradation. Comprehensive experiments across standard NLP benchmarks with heterogeneous LLM pools (7B-70B) demonstrate that GAR achieves substantial carbon reductions while maintaining competitive accuracy and p95 latency guarantees, providing a practical, theoretically grounded approach to sustainable LLM inference.

Disha Sheshanarayana, Rajat Subhra Pal, Manjira Sinha +1

Manipal University Jaipur · TCS Research

LLM Inference EfficiencyConstrained OptimizationLLM RoutingCarbon-Aware Computing
AskPDF
Feb 17, 2026

AI-CARE: Carbon-Aware Reporting Evaluation Metric for AI Models

As machine learning (ML) continues its rapid expansion, the environmental cost of model training and inference has become a critical societal concern. Existing benchmarks overwhelmingly focus on standard performance metrics such as accuracy, BLEU, or mAP, while largely ignoring energy consumption and carbon emissions. This single-objective evaluation paradigm is increasingly misaligned with the practical requirements of large-scale deployment, particularly in energy-constrained environments such as mobile devices, developing regions, and climate-aware enterprises. In this paper, we propose AI-CARE, an evaluation tool for reporting energy consumption, and carbon emissions of ML models. In addition, we introduce the carbon-performance tradeoff curve, an interpretable tool that visualizes the Pareto frontier between performance and carbon cost. We demonstrate, through theoretical analysis and empirical validation on representative ML workloads, that carbon-aware benchmarking changes the relative ranking of models and encourages architectures that are simultaneously accurate and environmentally responsible. Our proposal aims to shift the research community toward transparent, multi-objective evaluation and align ML progress with global sustainability goals. The tool and documentation are available at https://github.com/USD-AI-ResearchLab/ai-care.

KC Santosh, Srikanth Baride, Rodrigue Rizk

USD Artificial Intelligence Research, Department of Computer Science University of South Dakota Vermillion, SD 57069, USA

Carbon-Aware ComputingEnergy-Efficient MLMulti-Objective Optimization
AskPDF