Aman Amith Shastry Scientific Computation MCP server
{
"mcpServers": {
"smithery-aman-amith-shastry-scientific-computation": {
"type": "http",
"url": "https://server.smithery.ai/@Aman-Amith-Shastry/scientific_computation_mcp/mcp"
}
}
}{
"mcpServers": {
"smithery-aman-amith-shastry-scientific-computation": {
"type": "http",
"url": "https://server.smithery.ai/@Aman-Amith-Shastry/scientific_computation_mcp/mcp"
}
}
}{
"mcpServers": {
"smithery-aman-amith-shastry-scientific-computation": {
"type": "http",
"url": "https://server.smithery.ai/@Aman-Amith-Shastry/scientific_computation_mcp/mcp"
}
}
}{
"servers": {
"smithery-aman-amith-shastry-scientific-computation": {
"type": "http",
"url": "https://server.smithery.ai/@Aman-Amith-Shastry/scientific_computation_mcp/mcp"
}
}
}{
"mcpServers": {
"smithery-aman-amith-shastry-scientific-computation": {
"type": "http",
"url": "https://server.smithery.ai/@Aman-Amith-Shastry/scientific_computation_mcp/mcp"
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsonversion 1.13.1
What it does
An MCP server that exposes tools for linear algebra, vector calculus, and basic visualization. It maintains a per-session in-memory tensor store so that created matrices and vectors can be referenced by name across subsequent tool calls. Intended for developers or AI assistants that need programmatic scientific computation capabilities.
- Create matrices and vectors by name, then perform operations like matrix inversion, QR decomposition, or SVD on them within a session.
- Compute vector calculus operations such as gradient, curl, divergence, or directional derivatives on symbolic multivariable expressions.
- Generate 2D or 3D plots of scalar functions and 3D vector fields as PNG images.
Tools this server exposes
| Tool | Does |
|---|---|
| create_tensor | Creates a named tensor (vector or matrix) from provided shape and values and adds it to the session store. |
| view_tensor | Displays the contents of a named tensor from the session store. |
| delete_tensor | Removes a named tensor from the session store. |
| add_matrices | Adds two named matrices if their shapes are compatible. |
| subtract_matrices | Subtracts one named matrix from another if their shapes are compatible. |
| multiply_matrices | Multiplies two named matrices if their dimensions are compatible. |
| scale_matrix | Scales a named matrix by a scalar factor, in-place by default. |
| matrix_inverse | Computes the inverse of a named matrix. |
| transpose | Computes the transpose of a named matrix. |
| determinant | Computes the determinant of a named matrix. |
| rank | Computes the rank (number of pivots) of a named matrix. |
| compute_eigen | Computes the eigenvalues and eigenvectors of a named matrix. |
| qr_decompose | Computes the QR factorization of a named matrix. |
| svd_decompose | Computes the Singular Value Decomposition of a named matrix. |
| find_orthonormal_basis | Finds an orthonormal basis for the column space of a named matrix. |
| change_basis | Expresses a named matrix in a new basis. |
| vector_project | Projects one named vector onto another in the same vector space. |
| vector_dot_product | Computes the dot product of two named vectors. |
| vector_cross_product | Computes the cross product of two named vectors. |
| gradient | Computes the gradient of a symbolic multivariable scalar function. |
| curl | Computes the curl of a symbolic vector field given as a Python-list-formatted string. |
| divergence | Computes the divergence of a symbolic vector field given as a Python-list-formatted string. |
| laplacian | Computes the Laplacian of a scalar function or component-wise Laplacian of a vector field. |
| directional_deriv | Computes the directional derivative of a scalar function in a specified direction, normalizing the direction vector by default. |
| plot_vector_field | Returns a 2D PNG image of a 3D vector field plotted over configurable axis bounds. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.