NVSentinel Lab
Test NVIDIA NVSentinel on a kind lab: read its state and inject faults into a simulated GPU.
- tools
- 9
- transport
- STDIO
- env vars
- 3
- last updated
- Nov 9, 2026
Tools
9 toolsPublic tool metadata for what this MCP can expose to an agent.
lab_statusNodes (cordon, NVSentinel state, failing GPU checks), pods, the workload
Nodes (cordon, NVSentinel state, failing GPU checks), pods, the workload
No parameter schema in public metadata yet.
list_scenariosScenarios, stages, allowed XIDs, log sources
Scenarios, stages, allowed XIDs, log sources
No parameter schema in public metadata yet.
wait_for_stagePolls until a stage holds, such as `cordoned`
Polls until a stage holds, such as `cordoned`
No parameter schema in public metadata yet.
component_logsThe last lines of a NVSentinel component's logs
The last lines of a NVSentinel component's logs
No parameter schema in public metadata yet.
node_eventsRecent events on the GPU node
Recent events on the GPU node
No parameter schema in public metadata yet.
inject_faultInjects XID 95 into the simulated GPU
Injects XID 95 into the simulated GPU
No parameter schema in public metadata yet.
clear_faultRestarts the simulated hostengine, which drops the fault
Restarts the simulated hostengine, which drops the fault
No parameter schema in public metadata yet.
reset_labUndoes a DCGM outage and an injected fault
Undoes a DCGM outage and an injected fault
No parameter schema in public metadata yet.
run_test_scenarioRuns a scenario end to end and reports each step
Runs a scenario end to end and reports each step
No parameter schema in public metadata yet.
Sponsors
Agent CaffeineSponsoredClose your lid, come back to finished work
SnugSponsoredThe clipboard manager for the AI eraPromote your product hereReach 125k+ visitors monthly
WaxSponsoredBlock anything with supervision, without erasing your iPhonePromote your product hereReach 125k+ visitors monthly
knobsSponsoredSee your site on any phone, any day, in any languageConfiguration
NVSENTINEL_MCP_KUBECONFIGThe one kubeconfig it uses, private to you. Default in the image: /kube/config.
NVSENTINEL_MCP_ALLOW_CHANGESExactly true allows the change tools, on a kind cluster only. Off when unset.
NVSENTINEL_MCP_LAB_CONTEXTThe kubeconfig context changes need. Must start with kind-. Default kind-nvsentinel-lab.
Readme
nvidia-nvsentinel-mcp
An MCP server for testing NVIDIA NVSentinel (GPU fault detection and remediation on Kubernetes) on a lab cluster, written in Go like NVSentinel itself. An MCP client such as Claude Code can read what NVSentinel reports, inject a fault into a simulated GPU, and check that NVSentinel cordons and drains the node.
This repo holds the server and example test results. It does not install NVSentinel: that is omerfsen.aicr, on a cluster from omerfsen.nvke or on kind. It is meant to go upstream one day: see docs/upstream.md.
Test results
Real output from NVSentinel v1.30.0 on kind, through this server over MCP: docs/test-results. These prove NVSentinel's pipeline against simulated DCGM, not real GPU health.
Standards
This server follows these standards. Each requirement, its status, the test that proves it and the one known gap are in docs/mcp-standards.md, with links to check for newer versions.
Details
- Repository
- github.com/omerfsen/nvidia-nvsentinel-mcp
- Package
docker.io/omerfsen/nvidia-nvsentinel-mcp:0.5.1- Transport
- STDIO
- Registry
- MCP registry, active