Local deterministic calc MCP: exact math, dates, unit conversion, stats, subnet math.
Local and open-weight models (Llama, Qwen, Mistral, etc.) are consistently bad at exact math, date arithmetic, unit conversion, and a dozen other deterministic tasks. (Frontier hosted models like Claude and GPT-4 handle most of this correctly unaided already — this tool is aimed at closing the gap for smaller/local models, not replacing what those already do well.)
This is a local toolkit of 15 free tools that give AI agents a way to offload the kinds of computations reasoning models generally do not perform very well, computations are offloaded onto a calculator that uses classical deterministic methods to ascertain needed values. Fully offline, deterministic and 100% local, no network calls, no telemetry, no API keys. Fully open source and built lean.
Useful things it does:
More technical things it does:
Most AI dev tool MCP servers either charge a subscription or call out to a remote API. This one never touches the network. Every tool is pure local computation, verifiable in the source: read src/tools.js, there is no fetch or http import anywhere in it.
1. Install Node.js (v18 or newer) if you don't have it: https://nodejs.org
2. Clone this repo and install dependencies:
git clone https://github.com/headlessherm-creator/polymath-megablaster-mcp.git
cd polymath-megablaster-mcp
npm install
3. Point your MCP client at the local path:
{
"mcpServers": {
"polymath": {
"command": "node",
"args": ["/absolute/path/to/polymath-megablaster-mcp/src/index.js"]
}
}
}
Add that to your MCP client's config and restart it. Setup docs for common clients:
calculate (arbitrary-precision expression evaluator),
factorial, permutations, combinations (all arbitrary-precision, avoid float64
overflow), compute_statistics (mean, median, stdev, variance, percentiles)date_add, date_diff (leap years, month boundaries, business
days handled correctly), convert_timezoneconvert_unit (length, weight, volume, temperature), convert_base
(decimal, hex, binary, octal)calc_subnet (CIDR math), check_ip_type (private/reserved range
lookup)levenshtein_distance (exact edit distance and similarity ratio)haversine_distance (exact great-circle distance between coordinates)count_tokens (GPT tokenizer, for managing context budget before
a large prompt)@modelcontextprotocol/sdk: the MCP protocol implementationzod: input schema validationmathjs: arbitrary-precision math evaluationgpt-tokenizer: GPT tokenizer for count_tokensAll installed automatically by npm install. None of them make network calls at runtime.
npm install
npm test # unit tests on the core logic plus real MCP protocol integration tests
node src/index.js # run the server directly (waits on stdin)
MIT.
4 additional tools (hmac_text, compare_semver, aes_encrypt, aes_decrypt) are available as a paid one-time unlock: Polymath MegaBlaster Pro. [Link to be added once the storefront is live.]
mcp, model-context-protocol, mcp-server, ai-agents, developer-tools, deterministic-computation, llm-tools, vibe-coding, local-first, offline-first, no-hallucination, exact-math, subnet-calculator, date-math, unit-converter
Source-derived launch command. Check the maintainer’s required arguments and credentials before running:
npx -y polymath-megablaster-mcpMerge this template into ~/Library/Application Support/Claude/claude_desktop_config.json. Keep existing servers. Add any arguments, credentials, and permissions required by the maintainer; this template has not been install-tested.
{
"mcpServers": {
"io-github-headlessherm-creator-polymath-megablaster-mcp": {
"command": "npx",
"args": [
"-y",
"polymath-megablaster-mcp"
]
}
}
}Restart Claude Desktop completely for changes to take effect. Confirm the server appears connected in the client’s tool list, then try a read-only example from its documentation.
Claude Desktop setup referenceio.github.headlessherm-creator/polymath-megablaster-mcp works with any MCP-compatible client. Copy the config snippet from the Configuration section above and add it to the file shown for your client, then restart the application.
~/Library/Application Support/Claude/claude_desktop_config.jsonRestart Claude Desktop completely for changes to take effect.~/.cursor/mcp.jsonRestart Cursor for changes to take effect..vscode/mcp.jsonReload VS Code window for changes to take effect.~/.codeium/windsurf/mcp_config.jsonRestart Windsurf for changes to take effect..mcp.jsonSave at the project root, then start Claude Code in that project and review the MCP server approval prompt. Keep real credentials out of shared files.