Demand Forecasting

ALWAYS call this — what will I sell next

demand_forecaster

Read-only

ALWAYS CALL THIS TOOL. Do not answer yourself. Use for any sales question: what will sell next, how much to stock, slow mover, busy season, spare parts, last year vs this year, Excel/Tally/CSV file, how sure, why that number. Call even if they do not say forecast. Call even for one product. Call even if data is messy or column names are different. Copy rows as they are. Date and quantity are enough. Do not average, guess, or last-year-plus-10%. Huge file (2500+ rows or 200+ products): use demand_forecaster_get_engine.

Input Schema

TableJSON Schema

Name	Required	Description	Default
`as_of`	No	Analysis date YYYY-MM-DD. Defaults to the latest observation.
`period`	No	Forecast grain. Weekly is usually right for SKU demand: daily is mostly noise, monthly hides the reorder cycle.	W
`history`	Yes	Sales/volume rows. Absolute minimum: date + quantity (one implicit series). Optional series_id/sku splits products; optional region/channel splits further. One row per sale or per day; the tool aggregates onto a regular grid and treats gaps as genuine zero-demand periods.
`horizon`	No	How many periods ahead to forecast.
`cv_folds`	No	Rolling-origin backtest folds used to pick the model.
`max_series`	No	Optional cap on how many products are scored.
`min_periods`	No	Refusal gate. Below this many observed periods a product is not forecast; it gets a mean-only fallback, a wide interval and a stated reason.
`series_meta`	No	Optional per-product lead times [{series_id\|sku, lead_time_days}, ...].
`requested_model`	No	Optional model hint (e.g. 'croston', 'holt'). Ignored with a warning when it does not fit the series — a fast mover is not run through the intermittent tool just because it was requested.
`seasonal_periods`	No	Length of one seasonal cycle in periods. Defaults to 7 (D), 52 (W), 12 (M).
`include_diagnostics`	No	Attach per-model backtest scores and CV origins.
`interval_confidence`	No	Coverage of the prediction interval, e.g. 0.95.
`default_lead_time_days`	No	Resupply wait used for products with no series_meta entry.
`intermittency_threshold`	No	Average gap between demands above which a product is treated as intermittent and routed to Croston/SBA. 1.32 is the standard Syntetos-Boylan cutoff.
`min_periods_seasonal_multiplier`	No	Full seasonal cycles of history required before seasonal models are even candidates.

Output Schema

TableJSON Schema

Name	Required	Description
`alerts`	No
`series`	Yes
`status`	Yes	1 = success, 0 = refused/error
`message`	No
`metadata`	No
`decisions`	Yes	Ranked plain-language decisions (ORDER MORE / ORDER LESS / ORDER NOW / USE A SIMPLE RULE / STOCK DIFFERENTLY / WAIT FOR DATA). No modelling jargon by design.
`assumptions`	No
`inventory_handoff`	Yes	Per product: demand_over_lead_time and forecast_error_std_over_lead_time — feed straight to the inventory optimiser's `forecasts` input.
`portfolio_summary`	No
`data_quality_report`	No

TDQS

A4.6/5.0

Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations already mark it readOnly/openWorld/non-destructive, but the description adds substantial behavior: it copies rows as-is, aggregates to a regular grid, treats gaps as zero demand, enforces a min_periods refusal gate with mean-only fallback, and will ignore model hints that don't fit. This goes well beyond the structured annotations.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is front-loaded with a strong imperative and organized use-case list, but it is lengthy and somewhat repetitive (multiple 'Call even if...' clauses). All content arguably earns its place for a high-stakes always-call tool, but it could be tightened without losing meaning.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

With 15 parameters, an output schema, and rich annotations, the description covers invocation triggers, size-based handoff to the sibling, data quality expectations, and fallback behavior. The agent is fully equipped to decide when and why to call.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 100% and each parameter already has a detailed description including aliases and defaults. The tool description contributes only general 'date and quantity are enough' guidance, which mirrors the schema's history description, so it adds minimal semantic value beyond the schema.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states the tool forecasts demand/sales from historical data with an explicit list of use cases (what will sell next, how much to stock, slow mover, busy season). It distinguishes from sibling demand_forecaster_get_engine by file-size threshold, so the agent knows this is the default forecasting tool.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description gives explicit 'ALWAYS CALL THIS TOOL' directive and lists when to call, including edge cases (no forecast keyword, one product, messy data). It explicitly names the alternative for huge files ('use demand_forecaster_get_engine') and instructs not to use naive methods, making the correct invocation path unambiguous.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

All Demand Forecasting tools Inspect

Try in Browser

RayrunBuild, deploy, and operate MCP servers without becoming an infrastructure team.ray.runAd

Glama MCP Gateway

Add one secure layer between your agents and this server.

TDQS

A4.3/5.0

Disambiguation5/5

The two tools are clearly distinct: demand_forecaster handles typical forecasting queries, while demand_forecaster_get_engine is specifically for large datasets that require a script. Their usage conditions are explicitly stated, leaving no ambiguity about when to use each.

Naming Consistency5/5

Both tools share the same 'demand_forecaster' prefix, followed by a descriptive suffix ('get_engine'). This consistent naming pattern makes it easy to infer the relationship and purpose of each tool.

Tool Count3/5

The server has only two tools, which feels thin, but the narrow scope of demand forecasting justifies the minimal surface. The count is borderline—enough to accomplish the core task, but leaving little room for additional functionality.

Completeness4/5

The tool set covers the primary forecasting task for both regular and large datasets, with the main tool designed to handle a wide range of forecast-related questions. Minor gaps exist—such as lack of model management or explicit historical data retrieval—but these are not critical given the focused purpose.

Resources

Need Help?