Multi-Factory Energy & Resource Optimization Agent with Claude Agents SDK
High energy costs during peak tariff hours erode profit margins across multi-site manufacturing facilities. Build an energy optimization agent using Claude Agents SDK with extended thinking that dynamically re-indexes production schedules to match time-of-use electrical tariffs. Lower overall plant peak power costs by 15% without reducing daily production throughput.
What you are building
The core problem, expected build, and operating context for this challenge.
Develop a multi-site energy optimization agent using Claude Agents SDK to rebalance heavy machine power consumption across dynamic electricity tariff tiers.
How work is evaluated
Evaluates cost savings achieved through dynamic energy load shifting.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
peak_avoidance
Ensure job was not scheduled during peak price hour 14
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
energy_cost_saving
Percentage reduction in energy expenditure compared to baseline • target: 15 • range: 0-100
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Configure Claude Agents SDK extended thinking for evaluating multi-variable tariff matrices
Build function tools querying live electricity grid pricing and factory machine loads
Achieve 15% reduction in electricity cost while preserving required order throughput volumes
Reference links and supporting material
Hourly energy tariff profiles and heavy industrial machine power logs across 3 regional plants.
Challenge setup
Evaluates cost savings achieved through dynamic energy load shifting.
Challenge input
JSON machine power specs, job schedules, and hourly tariff schedule
Claude Agents SDK
Extended thinking capabilities facilitate complex trade-off reasoning across multi-hour dynamic windows.
Evaluated output
JSON rebalanced schedule with calculated total cost and peak demand kW
- Ensure job was not scheduled during peak price hour 14
- Percentage reduction in energy expenditure compared to baseline • target: 15 • range: 0-100
- Benchmark: Smart Factory Energy Optimization Suite (ENERGY-OPT-2025)
- Energy Cost Saving target: 15
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- Claude Code · Required
- Claude Code · Required
- Claude 4.1 Opus · Optional
- Claude 4 Sonnet · Optional
Evaluation contract
- peak_avoidance · Weight 1
- energy_cost_saving · Weight 1
Metrics
- energy_cost_saving · Target 15
Benchmarks
- Smart Factory Energy Optimization Suite (ENERGY-OPT-2025)
- energy
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON machine power specs, job schedules, and hourly tariff schedule
Output: JSON rebalanced schedule with calculated total cost and peak demand kW
Execution harness: Python
Evaluation tools
- Claude Agents SDK · Agent Framework
Recipe freeze marker
No recipe freeze time is recorded.