Use this local MCP server for deterministic calculations and developer utilities without network calls or API keys.
Math: arbitrary-precision expression evaluation, factorial, permutations, combinations, base conversion, and statistics.
Dates/time: date add/diff including business days, timestamp/date parsing, IANA timezone conversion, and cron next-run calculation.
Networking/geo: CIDR subnet details, IPv4 private/reserved range checks, and haversine distance.
Units/colors: convert length, weight, volume, temperature, and hex/rgb/hsl colors.
Text/agent utilities: GPT token counting, Levenshtein distance, line diffs, and UUID v4 generation.
Encoding/data: base64 encode/decode, hashing, JSON pretty/minify, JWT decode, and JSON/YAML/TOML config conversion.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Polymath MegaBlaster MCP Convert 5 miles to kilometers"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
A free, local, 16-tool MCP server that gives AI agents exact, deterministic answers for math, date arithmetic, unit conversion, subnet/networking calculations, statistics, token counting, and real random UUID generation — no network calls, no API keys, no telemetry.
If your AI agent is miscounting days between dates, rounding large numbers incorrectly, getting unit conversions slightly wrong, or messing up subnet/CIDR math, this MCP server exists specifically to fix that class of problem.
Local and open-weight models (Llama, Qwen, Mistral, etc.) are consistently unreliable at exact math, date arithmetic, unit conversion, and similar deterministic tasks — they "reason" their way to an answer instead of computing it, and get it subtly wrong. (Frontier hosted models like Claude and GPT-4 already handle most of this correctly unaided — this toolkit is aimed at closing the gap for smaller/local models, not replacing what those already do well.) Every tool here uses classical deterministic methods instead of model reasoning, so the answer is always exactly right.
Everyday things it fixes:
Exact math on huge numbers, no float rounding errors
Correct date math, including leap years and "last day of the month" edge cases
Unit conversion (km/miles, °C/°F, kg/lb) without approximation
Subnet/CIDR math for networking tasks
Great-circle distance between coordinates
More technical tools:
Arbitrary-precision arithmetic, factorials, permutations, combinations (no float64 overflow)
Statistics (mean, median, stdev, variance, percentiles) on a list of numbers
Base conversion (decimal, hex, binary, octal)
Levenshtein edit distance and similarity ratio (deterministic fuzzy matching)
IANA-aware time zone conversion
GPT tokenizer-based token counting, so an agent can manage its own context budget
Real cryptographically-random UUID (v4) generation — LLMs reliably fabricate plausible-looking but non-random, potentially colliding UUID strings instead
Related MCP server: Touchstone MCP Server
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 directly in the source: open src/tools.js — there is no fetch or http import anywhere in it.
1. Install Node.js (v18+) 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 install3. 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:
Claude Desktop: https://modelcontextprotocol.io/quickstart/user
Cursor: https://docs.cursor.com/context/model-context-protocol
Claude Code: https://docs.claude.com/en/docs/claude-code/mcp
Deterministic math: calculate (arbitrary-precision expression evaluator),
factorial, permutations, combinations (all arbitrary-precision, avoid float64
overflow), compute_statistics (mean, median, stdev, variance, percentiles)
Date and time: date_add, date_diff (leap years, month boundaries, business
days handled correctly), convert_timezone
Conversion: convert_unit (length, weight, volume, temperature), convert_base
(decimal, hex, binary, octal)
Networking: calc_subnet (CIDR math), check_ip_type (private/reserved range
lookup)
Text: levenshtein_distance (exact edit distance and similarity ratio)
Geo: haversine_distance (exact great-circle distance between coordinates)
Agent utility: count_tokens (GPT tokenizer, for managing context budget before
a large prompt), generate_uuid (real cryptographically-random v4 UUIDs, 1-100 at a time)
@modelcontextprotocol/sdk: the MCP protocol implementation
zod: input schema validation
mathjs: arbitrary-precision math evaluation
gpt-tokenizer: GPT tokenizer for count_tokens
All installed automatically by npm install. None of them make network calls at runtime.
deterministic math MCP server, MCP date calculator, MCP unit converter, subnet CIDR calculator MCP, local MCP server no API key, Claude MCP math tools, AI agent exact arithmetic, token counter MCP server.
Same local-first philosophy, different problems:
json-mender-mcp — fix malformed JSON, validate JSON Schema, test regex
hiring-mender-mcp — BambooHR + Greenhouse ATS write tools, one vocabulary
helpdesk-mender-mcp — Zendesk + Freshdesk + Help Scout write tools, one vocabulary
MIT
Decode a base64 string to text
Inputs: base64 (required).
Encode text as base64
Inputs: text (required).
Calculate subnet details (network, broadcast, usable hosts, mask) from a CIDR like 192.168.1.0/24
Inputs: cidr (required).
Evaluate an arbitrary-precision math expression exactly (fixes LLM multi-digit arithmetic errors). Supports +,-,*,/,^,sqrt(),sin(),log(), etc.
Inputs: expression (required).
Check if an IPv4 address is private/reserved and identify which range it falls in
Inputs: ip (required).
Compute nCr (combinations) exactly using arbitrary-precision integer math
Inputs: n (required), r (required).
Compute mean, median, stdev, variance, min/max, and percentiles (25/75/90) for a list of numbers
Inputs: numbers (required).
Convert a number between decimal, hex, binary, and octal
Inputs: value (required), toBase (required), fromBase (required).
Convert a color between hex, rgb(), and hsl() formats
Inputs: color (required).
Convert config data between JSON, YAML, and TOML
Inputs: to (required), from (required), input (required).
Convert a Unix timestamp (seconds or ms) to ISO 8601, or get the current time if omitted
Inputs: timestamp.
Convert a UTC/ISO date into a formatted date+time string for any IANA time zone
Inputs: date (required), timeZone (required).
Convert a value between units of length, weight, volume, or temperature (fixes LLM unit-math errors)
Inputs: value (required), toUnit (required), fromUnit (required).
Count tokens in text using the GPT-3.5/4 (cl100k_base) tokenizer. Useful for managing context budget before sending a large prompt
Inputs: text (required).
Add or subtract days/weeks/months/years/businessDays from a date, handling leap years and month boundaries correctly
Inputs: date (required), unit (required), amount (required).
Compute the exact difference between two dates in days, business days, and hours
Inputs: a (required), b (required).
Decode a JWT (header + payload + expiry check) without verifying the signature
Inputs: token (required).
Line-by-line diff of two text blocks, returning added/removed/unchanged lines
Inputs: a (required), b (required).
Compute n! exactly using arbitrary-precision integer math (avoids float64 overflow past ~18)
Inputs: n (required).
Pretty-print or minify a JSON string
Inputs: json (required), mode.
Generate one or more random UUID v4 values
Inputs: count.
Compute a hash digest of text (md5, sha1, sha256, sha384, sha512)
Inputs: text (required), algo.
Compute exact great-circle distance in km/miles between two lat/long coordinates
Inputs: lat1 (required), lat2 (required), lon1 (required), lon2 (required).
Compute exact edit distance and similarity ratio between two strings (deterministic fuzzy matching)
Inputs: a (required), b (required).
Parse a 5-field cron expression and compute the next N run times
Inputs: expression (required), count.
Parse a human/ISO date string into ISO 8601 and Unix timestamp forms
Inputs: date (required).
Compute nPr (permutations) exactly using arbitrary-precision integer math
Inputs: n (required), r (required).
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 27 tools
Each tool performs a clearly distinct utility operation; there is minimal overlap. The closest pairs are parse_date/convert_timestamp and diff_text/levenshtein_distance, but their descriptions differentiate them well.
Naming style is mixed and inconsistent. Some tools use verb_noun (format_json, hash_text), some use noun_verb (base64_encode), some are bare nouns (factorial, permutations), and others use unconventional order like date_add rather than add_date.
27 tools is large and more than needed for a coherent toolkit. The broad misc-utility scope inflates the count beyond the 3-15 well-scoped range, and many functions could be grouped into a smaller number of multi-purpose tools.
Covers many common utility categories like encoding, date math, math, and conversion, but the set is an arbitrary grab-bag. Obvious developer-utility gaps remain (URL/HTML encoding, string case manipulation, random generation, regex helpers), so coverage is useful but not comprehensive for a general-purpose toolkit.
500+ deterministic tools for AI agents: math, conversion, validation, hashing, encoding, date/time.
JSON/YAML, regex, diff, JWT, SQL dialects — the keyless millisecond ops an agent needs mid-task.
60+ units, live FX, timezones, and date arithmetic for AI agents.
Exact hashing, base64/hex/URL encoding, JWT decoding and UUIDs for AI agents. No auth required.
