Skip to main content
Glama

Related Servers

Alternatives to mcp-cuda4cpu

No user-submitted related servers found.

    Related Servers

    • A
      license
      A
      quality
      A
      maintenance
      eBPF-based GPU causal observability agent with MCP server. Traces CUDA Runtime and Driver APIs via kernel uprobes and host events via tracepoints to build causal chains explaining GPU latency. 7 tools: get_check, get_trace_stats, get_causal_chains, get_stacks, run_demo, get_test_report, run_sql. Telegraphic compression reduces token usage ~60%. Supports stdio and HTTPS (TLS 1.3) transport.
      11
      84
      -
    • A
      license
      A
      quality
      B
      maintenance
      Enables coding agents to profile and disassemble GPU kernels, benchmark and compare candidate rewrites against baselines, and run an autonomous speedrun loop that verifies and records results.
      10
      Apache 2.0
    • F
      license
      Not graded
      quality
      A
      maintenance
      Enables AI assistants to run GPU/CUDA environment diagnostics, check hardware and software compatibility, get safe install commands, and validate Dockerfiles for GPU issues through natural-language requests.
      -
    • F
      license
      Not graded
      quality
      A
      maintenance
      Enables coding agents to coordinate local profilers, benchmark tools, trace processors, and direct targets, turning explicit artifacts or commands into bounded runtime evidence with provenance, coverage, and limitations. It exposes analysis and capture tools that help trace, profile, and burn down hotspots in application and native code, GPU kernels, and inference stacks without workspace setup or a hosted service.
      -