Skip to main content
Glama

Related Servers

Alternatives to weight-editing-harness

No user-submitted related servers found.

    Related Servers

    • A
      license
      Not graded
      quality
      B
      maintenance
      Enables controlled access to language model internals via MCP, including activation capture, patching, ablation, steering, probe fitting, and bounded generation, with a persistent experiment ledger and human-approved compute.
      MIT
    • A
      license
      Not graded
      quality
      C
      maintenance
      Enables LLM clients to search, read, and patch a trusted source repository with rollback-safe edits, Git operations, and allowlisted validation over stdio or OAuth-protected HTTP. Keeps secrets and arbitrary shell access out of reach while enforcing bounded batched I/O and operator-controlled workspace trust.
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      MCP server for analyzing transformer language model internals, exposing tools to trace attention heads and neurons responsible for predictions, get activation statistics, ablate components, and sketch circuits.
      4
      37 PyPI
      1
      MIT
    • A
      license
      Not graded
      quality
      B
      maintenance
      Enables agents to turn a real model failure into a deterministic benchmark task by forging it with exact assertions, grading the task itself across six weighted factors, and inspecting a tamper-evident SHA-384 audit trail. Exposes eleven typed JSON-RPC 2.0 tools that call the same idempotent service layer as the UI, so mutations and decisions stay sealed and reproducible.
      MIT
    • A
      license
      Not graded
      quality
      B
      maintenance
      A universal interceptor that sits between any AI model's tool calls and shell execution, routing them through a supervisor daemon that inspects AST deltas, subprocess output, and hash-chained ledgers before work is allowed. It enables users to kill deleted assertions, trivial stubs, phantom test runs, and cyclic self-debate loops deterministically at the compiler, process, and filesystem boundaries rather than via prompt instructions.
      AGPL 3.0