Version 2.0 — Now with grounded RAG analysis

Forecasting your team can actually trust.

ChronoForge Pulse turns months of messy sales data into clear, defensible forecasts — and the production plans, material orders and analytics that follow. Built for finance and operations teams who'd rather decide than debate.

JLMRAK+
Trusted by 2,400+ planning teams SOC 2 Type II · GDPR · HIPAA
app.chronoforge.io / forecasts / q4-fy26
ForecastPlanRisk

Ensemble forecast — North region

SARIMA · LSTM · XGBoost · LightGBM, weighted by historical fit

Updated 8 minutes ago
Predicted revenue
$2.42M
+12.4% vs Q3
Confidence
0.82
Moderate
Ensemble accuracy
95.2%
90-day rolling
Risk factors
3
Review suggested
$2.42M Jul Aug Sep Oct Nov Dec
Try it · no signup

Ask Pulse anything.

A live snapshot of how the agent thinks — with the citations, the sparkline, the recommendation. Pick a question or write your own; the full workspace lives one click away.

Sample questions are pre-rendered. Free-text questions run live against Pulse — capped at 10 per visitor per day. The full agent is unmetered for signed-in users. Open the full agent →

Connects to the systems your team already runs
SAP Salesforce Snowflake Databricks Postgres NetSuite Power BI
Capabilities

Everything you need, nothing you don't.

An end-to-end planning workspace — from raw data to natural-language insight — without the usual stack of bolt-on tools.

Grounded answers, not guesses.

Ask a question in plain English. Pulse retrieves the historical scenarios that actually matter via vector search, then builds a cited, source-aware analysis with the LLM of your choice.

// natural language → grounded answer
POST /api/query
{
  "query":        "Why did Q3 sales soften in
                  the Pacific Northwest?",
  "region":       "PNW",
  "top_k":        5
}

// → 5 similar quarters retrieved
// → answer cites scenarios + variance

Four-model ensemble

SARIMA, LSTM, XGBoost and LightGBM run side-by-side. Each model's strengths cover the others' weaknesses — yielding 95.2% rolling accuracy.

Sub-200ms retrieval

HNSW-indexed cosine similarity over Postgres + pgvector. Millions of historical scenarios, all under a fifth of a second.

Risk intelligence

Variance bands, deviation factors and confidence scoring — automatically surfaced for every forecast you publish.

Choose your model

Claude, GPT, Mistral, Gemini — or run fully offline with local embeddings. Bring your own keys, swap any time.

Conversational planner

Multi-turn chat with your data. Drill into a region, slice by quarter, ask follow-ups — like having a senior analyst on call.

Architecture

From raw data to defensible decision.

Every query travels the same battle-tested RAG pipeline. No hidden behavior, no hallucinated numbers — just a transparent path from your data to a recommendation.

01

Ingest & embed

Sales scenarios, ERP exports and analyst notes are embedded with sentence-transformers and stored with rich metadata indexes.

02

Vector retrieval

Each query hits the HNSW index. The most contextually similar historical scenarios surface in milliseconds.

03

Context assembly

Retrieved scenarios are filtered by region, category and date, then formatted into a structured, auditable prompt.

04

Grounded generation

Your chosen LLM produces analysis with citations back to source rows — every claim is traceable to the data.

Ensemble engine

Four models. One prediction.

Each model brings a different perspective. The ensemble combines them — weighted by recent fit — into a forecast that's measurably more accurate than any individual.

SARIMA
Statistical · Time-series
91.8%
90-day rolling accuracy
LSTM
Neural · Sequence model
93.1%
90-day rolling accuracy
XGBoost
Gradient boosting
92.6%
90-day rolling accuracy
LightGBM
Gradient boosting
92.3%
90-day rolling accuracy
Ensemble accuracy: 95.2% ↑ 2.1% over best individual
Supply chain intelligence

Forecasting is just step one.

Pulse converts demand forecasts into production schedules, material plans and procurement orders — with optimization baked in. One workspace, one source of truth.

MPS

Master Production Schedule

Period-by-period production plans that balance capacity constraints against service-level targets. Chase or level, your call.

Available-to-promise Safety stock Capacity planning
MRP

Material Requirements

Multi-level BOM explosion. Calculates exactly what to order and when, with full lead-time offsetting across subassemblies and raw materials.

BOM explosion Lead-time offset Net requirements
LOT

Lot Sizing Optimizer

Five algorithms, from EOQ to Wagner-Whitin dynamic programming. Minimize total holding plus ordering cost — automatically.

Wagner-Whitin Silver-Meal EOQ / POQ
Pipeline output · run #4128
// ensemble forecast → MPS → MRP → lot sizing
{
  "total_cost":       $2,847,350,
  "service_level":    "94.2%",
  "planned_orders":   186,
  "bom_levels":       3,
  "best_lot_method":  "wagner_whitin",
  "cost_savings":     "12.4% vs lot-for-lot",
  "insights":         "3 paragraphs of grounded analysis →"
}
"We replaced a quarterly forecast workshop with a fifteen-minute review. Same accuracy, ten times less arguing."
M Maya Reyes · VP Demand Planning, Northwind Industrial
Developer experience

A REST API your team can read.

FastAPI-powered backend with auto-generated OpenAPI docs, typed schemas, and every endpoint your platform team needs to wire Pulse into production.

Endpoints
POST/api/querySingle RAG question
POST/api/chatMulti-turn analyst
POST/api/forecast/explainExplain a prediction
POST/api/forecast/riskRisk assessment
POST/api/forecast/recommendAction items
GET/api/kb/statsKB statistics
POST/api/supply-chain/runMPS → MRP → lot sizing
POST/api/supply-chain/analyzePlan + AI insights
forecast_explain.sh
# Explain a forecast with historical context
curl -X POST https://api.chronoforge.io/api/forecast/explain \
  -H "Authorization: Bearer $PULSE_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "date":                  "2026-12-15",
    "sarima_prediction":    245000.00,
    "lstm_prediction":      251200.00,
    "xgb_prediction":       248800.00,
    "lgb_prediction":       247500.00,
    "ensemble_prediction":  248125.00,
    "confidence_low":       232000.00,
    "confidence_high":      264000.00,
    "feature_importance": [
      { "name": "month",    "value": 0.34 },
      { "name": "promo",    "value": 0.28 }
    ]
  }'
Pricing

Two plans. No seat traps.

Full feature parity across plans. Start free, scale when you're ready, talk to us when you outgrow it.

Starter

For small planning teams

Everything you need to ship a trustworthy forecast — with room to grow.

$0/ month · 14 days
  • 4-model ensemble forecasting
  • Up to 50,000 historical scenarios
  • RAG-grounded chat with citations
  • Up to 5 seats
  • Email support
Start free — no card required
Most popular
Pro

For supply-chain teams

The full planning workspace — forecasting, MPS, MRP, lot sizing and analytics.

$399/ month
  • Everything in Starter
  • Full MPS → MRP → lot-sizing pipeline
  • Workforce / HR analytics module
  • Unlimited scenarios & chat sessions
  • Excel report export with charts
  • Priority support & onboarding
Start 14-day free trial

No card today · Cancel any time · Stripe checkout

Need SSO, custom contracts or on-prem? Talk to sales →

Plan with confidence.
Decide with data.

Spin up a workspace in under five minutes. No infrastructure. No PhD. Cancel anytime.