supplygraph-ai

Geographic Concentration Analysis Agent

Overview

The Geographic Concentration Analysis Agent performs a quantitative, multi-tier evaluation of an enterprise’s global supply chain to identify country-level and regional over-concentration risks.

By applying Herfindahl–Hirschman Index (HHI) calculations and deep-tier graph traversal, it exposes hidden dependency structures that are often invisible in traditional Tier-1 or Tier-2 analysis.

The agent delivers a clear, decision-ready report that highlights systemic vulnerabilities, enabling organizations to assess exposure, design mitigation strategies, and strengthen supply resilience before disruption occurs.

Pain

Most companies believe their supply chain is diversified — but in reality, hidden concentration often exists deep in Tier-3 through Tier-8 levels of their network.

Even globally diversified enterprises can rely on a single dominant geography for critical upstream inputs without realizing it. In one analysis, Nvidia — widely regarded as having a highly diversified supply chain — was found to have Tier-6 components with over 90% concentration in a single country based on HHI analysis.

Such invisible dependencies introduce severe enterprise risks:

Geopolitical shocks
Export controls and sanctions
Natural disasters
Regional instability

Any one of these can trigger cascading failures: component shortages, production delays, revenue loss, and reputational damage.

Breakthrough

Behind the scenes, SupplyGraph AI executes a node-by-node analysis of a company-centric supply graph, calculating geographic concentration at each tier using statistical benchmarks and real-world enterprise data.

User experience remains simple:

Provide a company name
Confirm the matched entity
Specify and confirm the target country or region
Receive a structured, fully documented concentration analysis

The final output includes:

HHI-based concentration metrics
Country and region dominance scores
Dependency distribution charts
Executive-ready risk summaries

Advanced network science, delivered through an intuitive and actionable format.

Why SupplyGraph AI

100M+ companies
1M+ key products
8,000+ industry benchmarks

SupplyGraph AI is the first AI-native infrastructure designed to perform deep-tier geographic dependency analysis at scale — without requiring any proprietary data uploads from clients.

Our platform transforms invisible structural risk into auditable, explainable intelligence that decision-makers can trust.

Try the Geographic Concentration Analysis Agent (Live Chatbot)

Before integrating this agent via API, you can experience it directly through our interactive chatbot.

This live demo allows you to:

Enter a company name
Confirm the correct entity
Specify a country or region for concentration analysis
Identify single-country dependency and hidden exposure risks
Preview the structure, clarity, and analytical depth of the final report

Launch the Geographic Concentration Analysis Agent
https://supplygraph.ai/zk_chat_os/agentic/dialog.html?name=sg_chokepoint

To use the chatbot, you’ll first need to:

Create a SupplyGraph AI account
Top up your credit balance

The chatbot is powered by the same Geographic Concentration Analysis Agent and A2A endpoints described in this documentation.
Credits consumed in the chatbot are deducted in the same way as API / A2A usage.

Everything you experience here can be fully embedded into your own systems through A2A integration.

Sandbox Key Support (for Development)

This agent supports Sandbox API Keys, enabling developers to test without consuming credits.

When using a Sandbox Key:

No credits are deducted
The agent returns a predefined geographic concentration report for Tesla, Inc. in China
Returned content follows the agent’s final output schema
No real computation or statistical modeling is executed under Sandbox mode
Only structural and format validation is performed

Suitable for:

Backend integration testing
Visualizing regional-risk heatmaps
Building analytics dashboards

⚠️ Sandbox datasets are static samples.
Production Keys are required for real, dynamic computations.

For more information, see
Getting Started Guide → API Keys.

Input Requirements

agent_id: sg_chokepoint · MCP tool: sg_chokepoint · Pricing: 264590 credits / run

Input differs by integration surface:

A2A & Agent API

Field	Required	Description
`text`	Yes	Natural-language input — company name, and later country or region when prompted

Example (initial): "Tesla, Inc."

The agent resolves the company and country/region through multi-turn confirmation before report generation starts (see Multi-turn Example (A2A)).

MCP

MCP does not use multi-turn confirmation on sg_chokepoint. Resolve pid and region_name via helper tools first.

Step	MCP tool	Input	Output
1	`search_company_candidates`	`text`	`candidates[].pid`
2	`search_region_candidates`	`text`	`candidates[].region_name`
3	`sg_chokepoint`	`pid`, `region_name`	Concentration report

Step 1 example: {"text": "Tesla United States"}
Step 2 example: {"text": "China"}
Step 3 example: {"pid": "a77828f060c866441f2403384b271e63", "region_name": "China"}

Helper tool pricing: 1 credit / run each for search_company_candidates and search_region_candidates.

Full machine-readable schemas → GET /api/v1/agents/sg_chokepoint/manifest (see Agent API §3).

Output Format

Primary output lives in data.content (Agent API) or task artifacts (A2A / MCP). The payload is a oneOf:

State	`content` type	Description
In progress, failed, cancelled, or waiting for user input	`string`	Text or Markdown — validation prompts, confirmation requests, or error messages
Task completed successfully	`object`	Structured geographic concentration report (see below)

Structured success object:

{
  "type": "sg_chokepoint_results",
  "data": [
    {
      "content_type": "dir",
      "content": "Summary"
    },
    {
      "content_type": "paragraph",
      "content": "### Research Objective\nThis report applies the Herfindahl–Hirschman Index (HHI) methodology to evaluate the geographic concentration and risk exposure of Tesla, Inc.'s global supply chain."
    }
  ]
}

Field	Description
`type`	Always `"sg_chokepoint_results"` on success
`data`	Ordered array of report elements
`data[].content_type`	`"dir"` — title or heading; `"paragraph"` — body text (may include Markdown)
`data[].content`	Text content of the element

Estimated task duration: 30 minutes – 7 hours (Production). Sandbox returns instantly.

After entity and region confirmation (A2A) or pid + region_name submission (MCP), the task enters a long-running executing state — poll status until TASK_COMPLETED, then call results.

Sample Response (Sandbox)

Sandbox returns the Tesla, Inc. / China fixture with the same structure as Production. Report content below is abbreviated.

Success (TASK_COMPLETED) — Agent API results:

{
  "success": true,
  "code": "TASK_COMPLETED",
  "message": "Task completed successfully.",
  "data": {
    "task_id": "<task-id>",
    "agent": "sg_chokepoint",
    "stage": "completed",
    "progress": 100,
    "content": {
      "type": "sg_chokepoint_results",
      "data": [
        {
          "content_type": "dir",
          "content": "Summary"
        },
        {
          "content_type": "paragraph",
          "content": "### Research Objective\nThis report applies the Herfindahl–Hirschman Index (HHI) methodology to evaluate the geographic concentration and risk exposure of Tesla, Inc.'s global supply chain in China."
        },
        {
          "content_type": "dir",
          "content": "HHI Analysis"
        },
        {
          "content_type": "paragraph",
          "content": "Tier-3 through Tier-6 components show elevated single-region dependency scores, with several nodes exceeding the 0.25 HHI threshold for high concentration."
        }
      ]
    }
  },
  "metadata": { "credits_used": 0 },
  "errors": null
}

Multi-turn Example (A2A)

A2A and Agent API only. MCP uses the three-tool workflow (search_company_candidates → search_region_candidates → sg_chokepoint) instead of multi-turn confirmation.

Each WAITING_USER response includes a Markdown prompt in content (string). Continue with the same task_id.

Turn 1 — submit company name:

curl -X POST "https://agent.supplygraph.ai/api/v1/agents/sg_chokepoint/run" \
  -H "Authorization: Bearer <YOUR_API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{"text": "Tesla, Inc.", "stream": false}'

Poll until WAITING_USER. Example content:

{
  "success": true,
  "code": "WAITING_USER",
  "data": {
    "task_id": "<task-id>",
    "content": "Here are the companies we've identified.\nCompany Name: Tesla, Inc.\nCountry: United States\nPlease reply [Yes] or [No] to confirm."
  }
}

Turn 2 — confirm company (same task_id):

curl -X POST "https://agent.supplygraph.ai/api/v1/agents/sg_chokepoint/run" \
  -H "Authorization: Bearer <YOUR_API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{"text": "Yes", "task_id": "<task-id>", "stream": false}'

Agent prompts for the target country or region.

Turn 3 — submit country or region (same task_id):

curl -X POST "https://agent.supplygraph.ai/api/v1/agents/sg_chokepoint/run" \
  -H "Authorization: Bearer <YOUR_API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{"text": "China", "task_id": "<task-id>", "stream": false}'

Poll until WAITING_USER with region confirmation, e.g.:

{
  "success": true,
  "code": "WAITING_USER",
  "data": {
    "task_id": "<task-id>",
    "content": "Country or region identified: **China**.\nPlease reply [Yes] or [No] to confirm."
  }
}

Turn 4 — confirm country or region (same task_id):

curl -X POST "https://agent.supplygraph.ai/api/v1/agents/sg_chokepoint/run" \
  -H "Authorization: Bearer <YOUR_API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{"text": "Yes", "task_id": "<task-id>", "stream": false}'

The agent queues report generation. Poll status until TASK_COMPLETED (may take 30 min – 7 h in Production), then retrieve sg_chokepoint_results via results.

See Agent API §8 for the general multi-turn pattern.

MCP Resolution Workflow

MCP integrations resolve entities upfront — no confirmation turns on sg_chokepoint.

Step 1 — `search_company_candidates`

{
  "type": "company_candidate_search",
  "data": {
    "candidates": [
      {
        "pid": "a77828f060c866441f2403384b271e63",
        "company_name": "Tesla, Inc.",
        "country_name": "United States"
      }
    ]
  }
}

Step 2 — `search_region_candidates`

{
  "type": "region_candidate_search",
  "data": {
    "candidates": [
      { "region_name": "China" }
    ]
  }
}

Step 3 — `sg_chokepoint`

report = await session.experimental.call_tool_as_task(
    "sg_chokepoint",
    {
        "pid": "a77828f060c866441f2403384b271e63",
        "region_name": "China",
    },
)
# poll until complete; parse sg_chokepoint_results

Integration

Method	ID / Tool	Documentation
A2A	`sg_chokepoint`	a2a.md
MCP	`search_company_candidates` + `search_region_candidates` + `sg_chokepoint`	mcp.md
Agent API	`sg_chokepoint`	agent-api.md

Call pattern is identical across agents — only agent_id / tool name and input differ.

Quick examples: A2A / MCP · Agent API

Errors

Common codes → Agent API §10.

Situation	Code	Agent behavior
Ambiguous or unmatched company (A2A / Agent API)	`WAITING_USER`	Prompts for entity confirmation; resume with same `task_id`
Country or region needs confirmation (A2A / Agent API)	`WAITING_USER`	Prompts for region confirmation after user supplies country/region name
Invalid `pid` or `region_name` (MCP)	`TASK_FAILED`	Returns guidance in `content` (string) — re-run helper tools
Insufficient credits	`INSUFFICIENT_CREDITS`	Top up via Console
Report still generating	`TASK_RUNNING`	Continue polling `status`; Production may take up to 7 hours

This site is open source. Improve this page.