# matrix-compute-mcp (slim)

> matrix-compute-mcp, an MCP server by matrix-compute.com, listed from the official MCP registry. Indexed, not reviewed: facts and our own checks, no score or ranking. P2P GPU matrix multiply, exact cosine search, and PCA/SVD verified via Freivalds' algorithm.

- Full: https://www.anchorterminal.com/tools/matrix-compute-mcp.md (~950 tokens) · this version ~880 tokens · JSON https://www.anchorterminal.com/tools/matrix-compute-mcp.json · canonical https://www.anchorterminal.com/tools/matrix-compute-mcp
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · MCP: https://www.anchorterminal.com/mcp · Updated: 2026-11-17
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)

# matrix-compute-mcp

> Indexed, not reviewed: facts from the official MCP registry and our own checks. No score, grade or rank, and not in the rankings until the panel reviews it. How the index works: https://www.anchorterminal.com/indexed/

- Kind: MCP server, by matrix-compute.com (https://matrix-compute.com)
- Category: GPU & serverless compute (https://www.anchorterminal.com/categories/gpu-compute.md)
- Listed because: It's published in the registry under matrix-compute.com, a namespace the registry only gives to whoever proves they control that domain.
- What the official MCP registry says: P2P GPU matrix multiply, exact cosine search, and PCA/SVD verified via Freivalds' algorithm.

## Facts

- MCP registry: `com.matrix-compute/matrix-compute-mcp` 0.1.4
- Endpoint: https://mcp.matrix-compute.net/mcp (streamable HTTP)
- Package: pypi `matrix-compute-mcp` (stdio)
- Source: https://github.com/matrixmultiplyconsortium/matrix-compute-mcp
- PyPI downloads a week: 414
- GitHub stars: 0
- Registry entry updated: 2026-11-15

## Tools

- Tools it lists (5, about 1,305 tokens of context, `tools/list` without credentials, checked 2026-11-17 04:18 UTC):
  - `get_network_status`: Queries the Matrix Compute discovery relay for live GPU worker availability, total VRAM capacity, confidential computing (TEE) enclaves, and sats_per_gflop…
  - `estimate_job_cost`: Calculates exact upload/download network bandwidth (MB), task count, GFLOPs, and satoshi cost estimates for a distributed matrix multiplication C (m x n) = A…
  - `execute_distributed_matmul`: Executes verified distributed matrix multiplication C = A x B across decentralized GPU workers with dynamic Freivalds verification and L402 settlement. Accepts…
  - `batch_similarity_search`: Runs exact cosine similarity search between query embeddings (M x D) and corpus embeddings (N x D) on distributed GPUs using OUTPUT_GRID_2D zero-reduction…
  - `compute_pca_svd`: Computes verified top-k Principal Component Analysis (PCA) / Singular Value Decomposition (SVD) on distributed GPU workers via randomized subspace iteration or…
- How its tools read to an agent (0 errors, 8 warnings, 1 note, about 1,305 tokens; rules at https://www.anchorterminal.com/check.md; not part of the score):
  - warn TC12 batch_similarity_search: no type for corpus_embeddings, queries
  - warn TC12 compute_pca_svd: no type for matrix
  - warn TC12 execute_distributed_matmul: no type for matrix_a, matrix_b
  - warn TC16 batch_similarity_search: no readOnlyHint or destructiveHint
  - warn TC16 compute_pca_svd: no readOnlyHint or destructiveHint
  - warn TC16 estimate_job_cost: no readOnlyHint or destructiveHint
  - warn TC16 execute_distributed_matmul: no readOnlyHint or destructiveHint
  - warn TC16 get_network_status: no readOnlyHint or destructiveHint
  - note TC24 server: 5 of 5 tools have no outputSchema

- JSON: https://www.anchorterminal.com/api/v1/tools/matrix-compute-mcp.json
- Being indexed says nothing about quality, and nobody can pay for it. Ask for a review: https://www.anchorterminal.com/builders/#claiming
