Install
{"mcpServers":{"calculate-mcp":{"command":"npx","args":["-y","calculate-mcp"]}}}Who It's For
- Developers who need high-precision integer arithmetic: built on native BigInt, 64-bit and 128-bit-plus integers with zero precision loss, supporting 8/16/32/64-bit signed and unsigned two's complement representations.
- People doing low-level debugging, reverse engineering, or protocol analysis: when it comes to bitwise operations, byte-order (endianness) swapping, or IEEE 754 floating-point layout decomposition, everything can be done in a single MCP server.
- People who need to chain multiple calculation steps into one execution: the batch_calc chained batch execution engine references previous step results with
{"$step": index, "field": "..."}, avoiding manual copying of intermediate values. - When it's NOT suitable: this is not a general-purpose chat or search tool; it runs on Node.js and requires an MCP client that supports stdio-mode configuration. The README does not mention it working in browsers or environments without Node — this is unconfirmed.
First Try in 3 Minutes
Prerequisite: Node.js is installed locally, and your MCP client supports the mcpServers configuration (the README provides this configuration format).
Recommended main path (labeled Option A, Recommended in the README): add the following to your client configuration:
{"mcpServers":{"calculate-mcp":{"command":"npx","args":["-y","calculate-mcp"]}}}
After saving, restart or refresh the client. The client will automatically fetch and start the server via npx (stdio mode) — no manual installation needed.
For the first call, it's recommended to pick a read-only, pure-computation capability, such as an arbitrary-precision integer conversion with int_convert, or decomposing a floating-point number with ieee754_convert — these are local numerical operations that involve no file writes or remote submissions.
Success signal: the tools provided by calculate-mcp appear in the client's tool list (the README describes a total of 29 dedicated tools), and a calculation request returns the correct result. If the tool list is empty, check that your client configuration format matches the one above.
Alternative method (for reference, not the main path): the README mentions you can also run it locally in stdio mode (npm start, after building the TypeScript into build/ first), which suits users who want to modify the source code or run offline.
Note: npx -y fetches and runs a package from npm, so the first execution downloads it over the network; this is the approach explicitly recommended by the official README.
Use Cases
Case 1: Verifying the bit-level representation of a 128-bit integer. Scenario: you're checking an integer larger than 64 bits within a protocol field. Call int_convert for an arbitrary-precision integer conversion, input your large integer and specify the target base/width; since the underlying implementation uses native BigInt, 64-bit and 128-bit-plus integers have zero precision loss, and 8/16/32/64-bit signed and unsigned two's complement representations are supported, you can directly compare the complement output against your expectations.
Case 2: Chaining multiple binary analysis steps in one go. Scenario: you need to perform bitwise operations on the same number, then swap its byte order, then do an IEEE 754 decomposition. Use batch_calc for chained batch execution: write all steps into a single batch call, with later steps referencing earlier results via {"$step": index, "field": "..."}. Expected result: a single call returns all intermediate and final results; the exact parameter names for each step should follow the tool's schema hints in your client — the README does not list the parameter details for every tool, and this part needs to be confirmed at runtime.
Usage Boundaries
- Read/write nature: all of the above tools perform local numerical/encoding-decoding computations; the README does not mention file writes, network submissions, or account operations, and offers no statement about an undo path, so no "reversible" claim can be made.
- Credentials: none; the configuration contains no keys or tokens.
- Runtime requirements: a Node.js environment; the implementation is pure TypeScript + Node.js standard library + Zod, with no heavy native binding dependencies.
- Explicit official limitations: the README provides no information on rate limits, permission boundaries, or data privacy — these key details are unconfirmed, and you should make no assumptions about privacy boundaries when using it.
- Still unconfirmed: the complete parameter list for each tool, operating system compatibility, and the specific version number of the npx package.