Use PullWeights from MCP-compatible AI tools — Claude Desktop, Cursor, Windsurf, and more. Available in TypeScript and Python.
TypeScript (npm)
npx @pullweights/mcp
Python (pip / uvx)
uvx pullweights-mcp # or pip install pullweights-mcp
Add to your claude_desktop_config.json:
{
"mcpServers": {
"pullweights": {
"command": "npx",
"args": ["-y", "@pullweights/mcp"],
"env": {
"PULLWEIGHTS_API_KEY": "pw_your_key_here"
}
}
}
}Add to your MCP settings (same format as above). Use uvx pullweights-mcp for the Python version.
PULLWEIGHTS_API_KEYAPI key (pw_... format). Required for pull, push, and listing orgs.PULLWEIGHTS_API_URLAPI base URL override (default: https://api.pullweights.com).Get your API key from the API keys dashboard.
Search for AI models by query, framework, or sort order.
query?string — Search query stringframework?string — Filter by framework (e.g. pytorch, gguf, safetensors)sort?string — Sort order: downloads, created, name, updatedper_page?number — Results per page (1-100, default 20)page?number — Page numberList models in an org, or list your orgs.
org?string — Organization name. Omit to list your orgs.Get the manifest for a model version — files, checksums, and metadata.
modelstring — Model reference: org/model or org/model:tagDownload model files to disk with SHA-256 verification.
modelstring — Model reference: org/model or org/model:tagoutput_dir?string — Output directory (default: ./pullweights_models/org/model/tag)Upload model files — two-phase: init, upload to S3, finalize.
modelstring — Model reference: org/model:tagfilesstring[] — Absolute file paths to uploaddescription?string — Model descriptionvisibility?string — public or private (default: public)Open source (MIT): github.com/pullweights/mcp