Batch Operations with AI Agents

Use MCP tools to deploy multiple GPU instances, run commands across a fleet, and check status in bulk — all through conversation.

Deploy a fleet of instances, run the same command on all of them, and monitor everything at once. Your AI agent handles the coordination.

---

”Deploy 4 RTX 4090 instances for a hyperparameter sweep.”

The agent provisions the fleet:

1. list_gpu_types — confirms RTX 4090 availability and pricing
2. create_instance — creates sweep-1 with RTX 4090
3. create_instance — creates sweep-2 with RTX 4090
4. create_instance — creates sweep-3 with RTX 4090
5. create_instance — creates sweep-4 with RTX 4090
6. instance_status — polls all four until running

Deployed 4 instances:

Instance	GPU	Status	Cost
sweep-1	RTX 4090	running	$0.35/hr
sweep-2	RTX 4090	running	$0.35/hr
sweep-3	RTX 4090	running	$0.35/hr
sweep-4	RTX 4090	running	$0.35/hr

Fleet burn rate: $1.40/hr. All 4 are ready for commands.

---

”Install dependencies on all of them and start training with different learning rates.”

The agent runs commands across the fleet:

1. ssh_execute on each instance — pip install torch transformers datasets
2. file_upload on each instance — uploads train.py
3. ssh_execute on each instance — starts training with different args:
   • sweep-1: python train.py --lr 1e-4
   • sweep-2: python train.py --lr 5e-5
   • sweep-3: python train.py --lr 1e-5
   • sweep-4: python train.py --lr 5e-6

All 4 instances are running training with different learning rates:

Instance	Learning Rate	Status
sweep-1	1e-4	training started
sweep-2	5e-5	training started
sweep-3	1e-5	training started
sweep-4	5e-6	training started

Check back in ~30 minutes for results, or ask me to monitor progress.

---

”Check training status on all instances and tear down any that are done.”

The agent checks each instance and cleans up:

1. ssh_execute on each — tail -5 /root/train.log
2. ssh_execute on each — nvidia-smi --query-gpu=utilization.gpu --format=csv,noheader

Instance	GPU Util	Last Log Line	Status
sweep-1	0%	Training complete. Best loss: 0.312	done
sweep-2	0%	Training complete. Best loss: 0.287	done
sweep-3	94%	Epoch 4/5 Step 800/1000 Loss: 0.301	running
sweep-4	91%	Epoch 3/5 Step 620/1000 Loss: 0.445	running

sweep-2 (lr=5e-5) has the best loss so far at 0.287. Terminating finished instances to save money.

3. file_download on sweep-1 and sweep-2 — downloads results
4. delete_instance — terminates sweep-1 and sweep-2

Downloaded results from sweep-1 and sweep-2. Terminated both. Fleet cost reduced to $0.70/hr (2 remaining).

---

Tools used in this workflow

Tool	Purpose
list_gpu_types	Check availability before bulk deployment
create_instance / instance_status	Deploy fleet and wait for readiness
file_upload / ssh_execute	Distribute code and run commands
file_download	Retrieve results from completed runs
delete_instance	Tear down finished instances