Deterministic Numerics.
Zero Credits. Zero Hallucinations.
Math is DataGrout's numeric engine for AI agents. Generate sequences, compute statistics, fit regressions, detect outliers, and normalize distributions — all in-process, deterministic, and completely free.
12 tools
generation · analysis · modeling
100% deterministic
seed for bit-for-bit reproducibility
Free forever
gateway costs waived entirely
Pairs naturally with Prism → to turn numeric output into charts and visual reports
Free to start · No credit card required
12 Tools Across Three Areas
Generation builds sequences, Analysis summarizes and cleans, Modeling fits and forecasts. All accept cache_ref — pipe data from any prior tool call without re-sending through the LLM.
math.range@1
math.range
Evenly-spaced sequences
Generate a numeric sequence from start to stop with a fixed step. Computed as start + i × step to avoid floating-point drift. Accepts negative steps for descending ranges. Maximum 10,000 values per call.
Key Parameters
Response Shape
What Agents Use Math For
LLMs can describe math. Math tools do it — deterministically, at scale, for free.
LLM-Free Descriptive Statistics
An agent receives 50,000 rows of sales data from Salesforce. Instead of sending values through the LLM for summarization, pipe the cache_ref to math.describe — get mean, median, percentiles, and a histogram in one call at zero cost.
Related Solutions
Anomaly Detection in Pipelines
Detect metric spikes, pricing anomalies, or sensor outliers before they propagate. math.outliers returns exact indices and bounds using IQR or z-score — no heuristics, no hallucinated thresholds.
Related Solutions
Regression & Trend Forecasting
Fit a linear or polynomial trend to time-series revenue data, get an equation string and r-squared, and extend N steps into the future. Agents can quote trend direction with reproducible numbers — not LLM guesses.
Related Solutions
Reproducible Synthetic Datasets
Seed math.sample with an integer and get bit-for-bit identical draws every run. Build test fixtures, simulate load patterns, or generate benchmark datasets without touching production data.
Related Solutions
Correlation Analysis Across Fields
Pass a Frame tool result via cache_ref to math.correlate — get a full Pearson and Spearman matrix with r-squared and interpretation labels across every field pair. No LLM inference, instant results.
Related Solutions
Easing Curves for Animations & UI
Generate a linspace, pipe it through math.interpolate with ease_in_out_cubic or any of 12 easing functions, and pass records to prism.chart. Build smooth animation keyframes or progress curves without writing any math.
Related Solutions
Math Gets More Powerful With Other DataGrout Tools
Chain Math into pipelines with Prism, Frame, Flow, and Logic to build complete numeric intelligence workflows.
Instant Visualization Pipeline
Pass Math output records directly to prism.chart via cache_ref. Generate a histogram from math.describe, plot a regression fit from math.trend, or visualize an easing curve from math.interpolate — all in two tool calls.
✦ Numeric output becomes a hosted chart in one hop
Pre-Filter Then Analyze
Use frame.filter to isolate a relevant slice of records (e.g. Q4 revenue only), then pipe the cache_ref into math.describe or math.trend. No re-fetching, no LLM overhead — the data stays server-side throughout.
✦ Sub-second statistical analysis on pre-filtered datasets
Automated Analysis Workflows
Build a flow that fetches data from an integration, runs math.outliers to detect anomalies, branches on findings with flow.route, and notifies via a downstream tool — all scheduled via Governor.
✦ Autonomous anomaly detection without manual intervention
Persist Numeric Facts
Assert math results as Logic facts — correlation coefficients, trend directions, outlier counts. Future agent calls can query these without re-running the analysis, making historical patterns available at sub-millisecond speed.
✦ Numeric intelligence that persists across agent sessions
Math in Practice
One import. Auth, retries, and cache_ref chaining all handled by the Conduit SDK.
from datagrout.conduit import Client
async with Client(
"https://gateway.datagrout.ai/servers/{uuid}/mcp",
auth={"bearer": "your-access-token"}
) as client:
# Step 1: fetch data from Salesforce (result cached server-side)
sf_result = await client.perform(
"salesforce@1/get_opportunities@1", {}
)
cache_ref = sf_result["_meta"]["datagrout"]["cache_ref"]
# Step 2: describe the 'amount' field — no data re-sent
stats = await client.perform("data-grout@1/math.describe@1", {
"cache_ref": cache_ref,
"field": "amount",
"bins": 20
})
print(f"Mean: {stats['mean']:.2f}")
print(f"Std: {stats['std']:.2f}")
print(f"P95: {stats['percentiles']['p95']:.2f}")
# stats['histogram'] → ready for prism.chartAlso available in TypeScript, Rust, Elixir, and Ruby. View Conduit SDK →
Frequently Asked Questions
Everything developers ask before integrating Math tools.
More questions? Read the full Math documentation.
Deterministic numerics. Zero credits. Production-ready.
Math tools run fully in-process — no LLM, no external calls, no token cost. Chain them with Prism, Frame, and Logic to build complete numeric intelligence pipelines for your agents.
Free to start · No credit card required