A real-time, agentic cost intelligence system powered by the Microsoft Agent Framework, built to compute and visualize AI project costs dynamically using live cloud pricing and autonomous reasoning.
AICalcX provides real-time, transparent, and intelligent cost estimation for AI projects — combining live pricing APIs, human resource cost inputs, and autonomous agents for multi-dimensional cost breakdowns.
Unlike static calculators, AICalcX adapts dynamically to live Azure pricing, usage patterns, and AI model parameters, producing accurate TCO (Total Cost of Ownership) for enterprise-scale AI workloads.
- Live Pricing Fetch — Azure Retail API + SerpAPI fallback
- Multi-Agent Orchestration — modular agents for pricing, infra, AI, ops, and human cost
- Natural Language Queries — e.g., “Cost of 1M GPT-4 tokens in eastus”
- Dynamic Token Computation — understands shorthand (e.g., “500k”, “2.5B”)
- Cross-Cloud Ready — Azure, AWS, GCP, Render, Vercel (upcoming)
- Transparent Reports — detailed cost breakdowns with live references
- Scenario Simulation — scale users, agents, or tokens dynamically
- No Hardcoded Pricing — all fetched from APIs or MCP-managed data
| Agent | Role |
|---|---|
| IntentAgent | Interprets user queries and extracts model, token count, and region. |
| PricingAgent | Fetches real-time cloud pricing via Azure Retail API or MCP cache. |
| FallbackAgent | Uses SerpAPI or cached JSON when APIs fail or data is missing. |
| AIAgent | Calculates AI model usage cost (tokens, embeddings, fine-tuning). |
| InfraAgent | Estimates compute, storage, and network infrastructure costs. |
| HumanAgent | Handles human resource cost — client-side and vendor-side personnel. |
| OpsAgent | Computes operational overhead (maintenance, monitoring, contingency). |
| AggregatorAgent | Consolidates all agent outputs into unified markdown tables. |
User Query
↓
IntentAgent → PricingAgent → Azure Retail API
↓ ↑
FallbackAgent ← SerpAPI
↓
AIAgent / InfraAgent / HumanAgent / OpsAgent
↓
AggregatorAgent → Markdown / CSV Report Output
Each agent operates autonomously, fetching real-time data and updating a shared AgentContext before aggregation.
Prompt:
“Estimate cost of 1M GPT-4 input tokens and 500k output tokens in East US.”
Result:
| Type | Tokens | Unit Price | Formula | Cost |
|---|---|---|---|---|
| Input | 1,000,000 | $0.0022/1K | (1M ÷ 1K) × 0.0022 | $2.20 |
| Output | 500,000 | $0.0300/1K | (500k ÷ 1K) × 0.03 | $15.00 |
| Total | — | — | — | $17.20 |
- User Input
- Fill project parameters via Chainlit or natural language query.
- Agent Execution
- Each agent fetches or computes its domain-specific costs.
- Cost Aggregation
- AggregatorAgent combines all outputs into a single structured report.
- Visualization & Export
- Display breakdowns or export as CSV/JSON/PDF.
Cost of Standard_D4s_v3 in eastusEstimate 2M GPT-4 tokens in japaneastAzure SQL Managed Instance DSU1500 costRender GPU instance 8GB VRAM
# Clone the repository
git clone https://github.com/ShivamGoyal03/AICalcX.git
cd AICalcX
# (Optional) Create a virtual environment
python -m venv .venv
source .venv/bin/activate # or .venv\Scripts\activate on Windows
# Install dependencies
pip install -r requirements.txt
# Run the app
python app.py- AWS / GCP / Vercel pricing integration
- Visualization dashboard
- Autonomous reasoning loop
- Budget alerting and cost tracking
- REST / GraphQL API support
Refer to BLOCKERS.md for full context.
Summary of major challenges:
- AI Model Pricing Complexity – multiple SKUs, dynamic regions, unclear mapping.
- Human Resource Variability – non-standardized rates and team compositions.
- Deployment Costs – regional pricing differences and scaling complexity.
- Usage Estimation Uncertainty – unpredictable token consumption.
- Azure Pricing API Gaps – delayed updates and inconsistent SKUs.
- Subscription Variations – enterprise vs. retail pricing mismatches.
- Cloud Type Differences – public, government, and sovereign region issues.
These limitations led to temporary project suspension pending reliable, unified data mapping.
This project is licensed under the MIT License.
See the LICENSE file for details.
Contributions are welcome!
If you'd like to improve agents, integrate new cloud providers, or enhance pricing logic, fork the repository and submit a pull request.
Note: This repository currently serves as a Proof of Concept (PoC) for open-source release, demonstrating how autonomous agents can perform real-time AI project cost computation using live cloud data and MCP-based orchestration.