All Services

Computes the correlation between completed_tasks and assigned_tasks, plus a weighted mean of a chosen secondary measure, across up to 200 research task completion records grouped by company and date.

Returns deterministic, byte-exact JSON metadata for the npm package 'left-pad' via a paid artifact gateway endpoint.

Returns a paid agent-facing decision procedure brief on how to handle reservation deposits and cancellation policies when booking on behalf of users

Returns an AI-generated decision procedure (prose brief) describing how an autonomous agent should handle the full lifecycle of reservation hold settlement — from hold creation through settlement or expiry.

Searches recent Google News RSS coverage of retail and ecommerce topics tailored for developers, returning publisher links and publication dates.

Searches recent Google News RSS for retail and ecommerce coverage relevant to operations teams, returning publisher links and publication dates.

Searches recent Google News RSS feeds for retail and ecommerce coverage specifically relevant to technology buyers, returning publisher links and publication dates.

Calculates the commission amount and remaining sale amount after commission deduction, given a sale amount and commission percentage, tailored for retail workflows.

Calculates the discounted price and savings amount for a retail item given an original price and a discount percentage.

Calculates a retail invoice subtotal from line items, applies a user-supplied tax rate, and returns the total amount due.

Analyzes up to 500 seller-supplied CSV orders to flag missed shipping/delivery deadlines and inconsistent status timestamps, returning results in JSON and CSV.

Calculates the absolute difference and percentage change between a start and end value, tailored for retail workflows.

Searches live public web pages about retail point-of-sale integrations, returning source links and snippets tailored for analyst-level research.

Calculates MRR, ARR, and expected monthly churn for retail subscription businesses from subscriber count, monthly price, and churn rate inputs.

Compares complete, partial, and missed evidence recovery across a matched, hash-frozen cohort of up to 64 question-paper pairs, reporting paired transitions and a caller-declared complete-recovery drop threshold.

Calculates the value recovery rate for returned goods across up to 200 retail records in a single batch call.

Computes a benchmark correlation report comparing recovered value to returned goods cost across supplied retail records.

Bins up to 200 returned-goods value recovery observations into a distribution histogram with configurable bin count

Computes a statistical distribution summary of recovery rates for returned goods given pairs of recovered value and original returned goods cost.

Compares named what-if scenarios against a baseline to evaluate recovered value rates for returned retail goods, returning a bounded report per call.

Analyzes ordered trend patterns in returned goods value recovery ratios from supplied retail row data (up to 200 records).

Ranks retail return records by value recovery rate, accepting up to 200 rows of recovered value and returned goods cost data.

Computes a rolling mean of recovered value and returned goods cost metrics over a configurable window size, returning a smoothed time-series report for retail return recovery analysis.

Computes a sequential (waterfall) attribution of how changes in returned-goods cost and recovered value contribute to shifts in a retail return recovery metric, comparing a 'before' and 'after' scenario.

Calculates the gap between actual recovered value from returned goods and a user-supplied target, based on returned goods cost and recovered value inputs.

Validates a CSV file of return import data and reports any blank rows found in it.

Validates calendar date fields in a return-imports CSV file, checking up to 1000 rows for date correctness and consistency.

Validates CSV cell lengths in return import data against configurable maximum character limits, supporting up to 1000 rows.

Validates a CSV table of return import data against a declared column contract, checking up to 1000 rows for structural and value conformance.

Validates a CSV of return import records and exports a report identifying all invalid rows that violate declared constraints.

Validates a CSV table of return import records and returns a located error report identifying specific cell-level errors

Validates that all numeric columns in a return-imports CSV table contain valid numeric values, within optional min/max bounds and allowed categorical values.

Validates a CSV table of return import records and flags spreadsheet formula injection risks across up to 1000 rows.

Validates a CSV file of return import data against a declared table contract and exports only the valid rows, up to 1000 rows.

Validates a CSV table of return import records for whitespace anomalies, checking up to 1000 rows with support for quoted cells and configurable delimiters.

Calculates Return on Average Assets (ROAA) from supplied figures and measures the result against a user-provided numeric target, returning the gap between actual and target values.

Calculates a weighted mean of Return on Average Assets (ROAA) across up to 200 supplied data rows using user-provided nonnegative weights.

Rebases retail return processing cost-per-case figures to a base-100 index, normalizing labour, postage, and other costs across up to 200 data rows.

Computes and compares the return processing cost per case between two time periods using supplied labour, postage, and other costs alongside processed return volume.

Computes benchmark correlation between retail return processing cost per case (labour, postage, other costs) and supplied benchmark values across up to 200 records.

Computes a distribution summary (min, max, mean, median, percentiles, etc.) for the return processing cost per case metric across up to 200 supplied retail records.

Computes IQR-based outlier flags for return processing cost per case across supplied retail records (up to 200 rows).

Computes the local finite-change elasticity of return processing cost per case with respect to a specified input variable (labour, postage, or other cost, or processed returns volume) at a given baseline.

Computes a one-input sensitivity table for retail return processing cost per case by varying one cost driver (labour, postage, other cost, or processed returns) across specified values while holding the rest at baseline.

Performs an ordered trend analysis on retail return processing cost-per-case data, decomposing labour, postage, and other costs across up to 200 time-ordered records.

Ranks up to 200 retail records by their return processing cost per case, computed from labour, postage, and other costs divided by processed returns.

Calculates the gap between actual return processing cost per case and a user-supplied target, using retail cost inputs (labour, postage, other costs, and volume).

Classifies retail return processing cost per case as below, within, or above a defined numeric threshold range, given labour, postage, and other cost inputs.

Generates a fixed-holdout learning curve for a naive-Bayes return-reason text classifier using up to 200 user-supplied labelled records, without any hosted or foundation-model training.

Runs a grouped three-fold cross-validation baseline evaluation on up to 200 labelled return-reason text records using a naive-Bayes classifier, returning accuracy and fold metrics.

Audits up to 200 labelled text records for return-reason classification accuracy, computing evaluation metrics against actual vs. predicted labels across four categories.

Computes label distribution statistics across up to 200 labelled return-reason text records, categorized as damaged, wrong_item, unwanted, or other.

Exports up to 200 labelled return-reason text records in JSONL format for local ML evaluation and naive-Bayes baseline training

Computes macro, micro, and weighted precision/recall/F1 aggregate scores for a return-reason text classification dataset with up to 200 labelled records.

Computes the multiclass Brier score for a return-reason classifier given up to 200 labelled text records with predicted probabilities.

Computes multiclass log loss for return-reason text classification models across up to 200 labelled records

Computes per-label precision and recall metrics for a return-reason text classifier across up to 200 labelled records with actual and predicted labels.

Evaluates a supplied set of labeled return-reason text records (up to 200) broken down by group, computing per-group classification metrics against predicted and actual labels.

Generates a taxonomy prompt template for classifying product return reasons (damaged, wrong_item, unwanted, other) from up to 200 labelled text records

Computes a bounded local finite-change elasticity report showing how small percentage changes in one retail return/restock variable affect the net restock value after processing.

Calculates the net restock value recovered from returned retail units after processing costs, given unit inventory value and a recoverable percentage.

Searches live public web pages for information about returns management platforms, returning source links and snippets relevant to agent builders.

Searches live public web pages for information about returns management platforms targeted at developers, returning source links and snippets.

Searches live public web pages for information about returns management platforms tailored to operations teams, returning source links and snippets.

Searches live public web pages for returns management platform information specifically relevant to technology industry buyers, returning source links and snippets.

Validates whether a submitted deliverable meets a given specification for a funded task, returning a deterministic true/false verdict

Lists funded task opportunities with optional filtering by count and minimum reward amount

Verifies whether a given URL represents a funded task with valid, deterministic deliverable validation

Determines whether a funded task or deliverable is eligible to be claimed, given a URL and canonical key

Audits a website's call-to-action path, returning evidence of issues and a concrete repair action for $0.09 USDC.

Runs a paid mobile markup diagnostic on a given domain, returning evidence of issues and a concrete repair action for $0.07 USDC.

Runs a paid self-serve diagnostic crawl on a given domain and returns actionable evidence-backed repair recommendations

Audits a website domain for broken or problematic links and returns evidence plus a concrete repair action for $0.10 USDC.

Calculates adjacent period-over-period changes in revenue per visitor from supplied retail row data.

Calculates revenue per visitor (RPV) for a batch of up to 200 row pairs of revenue and visitor counts supplied by the caller.

Calculates the correlation between revenue-per-visitor figures and supplied benchmark values for retail scenario analysis.

Computes a distribution histogram of revenue-per-visitor metrics from supplied retail row data, bucketed into configurable bins.

Calculates the local finite-change elasticity of revenue per visitor for a retail scenario given baseline revenue and visitor figures and a step size.

Compares multiple named scenarios for revenue-per-visitor metric using supplied retail figures, returning scenario-by-scenario analysis locally without live data access.

Performs ordered trend analysis on a supplied time-series of revenue and visitor figures to surface directional patterns in revenue per visitor (RPV).

Ranks a list of supplied revenue/visitor record pairs by their computed revenue-per-visitor metric.

Calculates revenue per visitor (RPV) metric from supplied total revenue and visitor count figures

Calculates the gap between current revenue-per-visitor and a specified target, enabling retail scenario analysis.

Classifies revenue-per-visitor figures as below, within, or above a defined threshold range using supplied retail data.

Generates a two-dimensional scenario grid showing revenue-per-visitor outcomes across combinations of revenue and visitor count values.

Computes the weighted mean of revenue-per-visitor across multiple retail data rows, using caller-supplied revenue and visitor counts with optional per-row weights.

Validates x402 USDC payment compatibility for a Base wallet or endpoint before committing a pay-per-call transaction

Computes confidence calibration bins for a review aspect classifier, evaluating how well predicted probabilities match actual outcomes across performance, appearance, value, and other categories.

Computes a confusion matrix for a review aspect classifier (performance, appearance, value, other) given up to 200 labelled text records with actual and predicted labels.

Generates a fixed-holdout learning curve for a naive-Bayes review aspect classifier (performance/appearance/value/other) from up to 200 labelled text records.

Audits classifier labels and predictions for review aspect classification (performance, appearance, value, other) across up to 200 labelled text records, computing evaluation metrics using naive-Bayes baselines.

Exports up to 200 labelled review-aspect text records as a JSONL file, with each record tagged as performance, appearance, value, or other.

Computes macro, micro, and weighted aggregate classification scores for a review aspect classifier (performance, appearance, value, other) given up to 200 labelled text records with actual and predicted labels.

Computes the multiclass Brier score to evaluate probabilistic classifier quality on review aspect labels (performance, appearance, value, other)

Computes multiclass log loss metric for review aspect classification (performance, appearance, value, other) from up to 200 labelled text records with predicted probabilities.

Generates a taxonomy prompt template for classifying review text into aspects (performance, appearance, value, other) using up to 200 labelled training records.

Computes top-k label coverage metric for a review aspect classifier given up to 200 labelled records with predicted probabilities.

Checks whether supplied evaluation texts overlap with labelled training records used to train a local naive-Bayes review-aspect classifier, detecting data leakage.

Trains a local naive-Bayes text classifier on up to 200 labelled review records and classifies query texts into review aspect categories (performance, appearance, value, other).

Compares a revision transition against its declared predecessor to detect lineage inconsistencies and return an integrity receipt