The Corporate Exception Report Agent continuously monitors and analyzes companies to detect critical exceptions, risk signals, regulatory events, and abnormal changes that can impact operational stability, compliance posture, or reputation.
It transforms fragmented, unstructured external information into a structured, verified, and continuously updated intelligence report, enabling organizations to respond to corporate risks in near real-time.
Traditional corporate monitoring is slow, fragmented, and resource-intensive.
Analysts frequently spend 10+ hours per company gathering information from scattered public sources — and by the time a report is completed, much of the data is already outdated.
Most organizations also lack:
This leads to blind spots in:
The Corporate Exception Report Agent automates what previously required multiple teams and systems.
Behind the scenes, it:
For the user, the workflow remains simple:
Typical insights include:
Powered by:
SupplyGraph AI provides a continuously updated, audit-ready corporate exception intelligence layer without requiring any proprietary data.
This makes the agent suitable for:
This agent fully supports Sandbox API Keys, enabling developers to test integrations without consuming credits.
When using a Sandbox Key:
Sandbox Keys are ideal for:
⚠️ Sandbox data must not be used for production analytics.
The full A2A workflow (run → status → results) behaves the same as production, enabling accurate integration testing.
When a user initiates a new monitoring request (via /run without a task_id), the agent begins constructing a full baseline dataset for the specified company.
This includes:
This initial build requires approximately 2–3 hours.
First-time workflow:
/run without task_id → system creates task_idTASK_COMPLETED is returned → first dataset is readymode=resultsAfter this initial build:
When a user confirms monitoring for a specific company:
resultsA subscription remains active as long as:
A task_id represents a subscription instance, not a company identity.
/run without task_id → system generates a new oneImplications:
/run, the system treats it as a new subscription, even for the same companyThe results endpoint always returns the current latest snapshot for the monitored company, as long as the task_id is active.
status every 10 secondsstatusresults to fetch the latest updated snapshotA subscription may end through:
Users may cancel by calling /run with mode=cancel + task_id.
curl -X POST https://agent.supplygraph.ai/api/v1/agents/corporate_exception_report/run \
-H "Authorization: Bearer <YOUR_API_KEY>" \
-H "Content-Type: application/json" \
-d '{
"mode": "cancel",
"task_id": "<task-id>"
}'
{
"success": true,
"code": "TASK_CANCELLED",
"data": {
"task_id": "<task-id>",
"agent": "corporate_exception_report",
"stage": "cancelled",
"content": "Your subscription has been successfully cancelled. Monitoring will continue until the current billing period ends."
}
}
If credits cannot be deducted:
results/status return no meaningful dataAny restart requires a new subscription, resulting in a new task_id.
agent_id: corporate_exception_report · MCP tool: corporate_exception_report
| Field | Required | Description |
|---|---|---|
text |
Yes | Company name and country |
chapter_name |
No | Exception domain; default ALL |
chapter_name values: ALL, Profile, License, GovRel, R&D, Reputation, Ops, GeneralRisks, Cost, Compliance, Competition, Brand, HR
Supported modes: run, status (initial build only), results, cancel. See Agent Behavior Model above for subscription lifecycle.
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, subscription inactive, or waiting for user input | string |
Text or Markdown — validation prompts, confirmation requests, or error messages |
| Task completed successfully | object |
Structured corporate exception report (see below) |
Structured success object:
{
"type": "results",
"data": {
"languages": ["en", "zk"],
"report_infos": {
"en": {
"report_date": "2025-11-30",
"report_name": "Corporate Exception Report — Tesla, Inc.",
"report_type": "enterprise changes report"
},
"zh": {
"report_date": "2025-11-30",
"report_name": "企业异常报告 — Tesla, Inc.",
"report_type": "enterprise changes report"
}
},
"chapter_infos": [
{
"en": {
"chapter_name": "Compliance",
"sub_sections": [],
"chapter_texts": ["Markdown paragraph…"]
},
"zh": {
"chapter_name": "合规风险",
"sub_sections": [],
"chapter_texts": ["Markdown 段落…"]
}
}
]
}
}
| Field | Description |
|---|---|
type |
Always "results" on success |
data.languages |
Language codes in the response — en (English), zk (Chinese) |
data.report_infos |
Report metadata keyed by language (en required; zh when Chinese is included) |
data.chapter_infos[] |
Array of exception-domain chapters, each with en / zh blocks |
chapter_infos[].en\|zh.chapter_name |
Chapter title (e.g. Compliance, Reputation, Ops) |
chapter_infos[].en\|zh.sub_sections[] |
Nested sub-sections (section_title, sub_sections, section_texts) |
chapter_infos[].en\|zh.chapter_texts[] |
Chapter body paragraphs (Markdown) |
During an active subscription, each results call returns the latest snapshot of the monitored company — not a one-time static export.
Initial baseline build: ~2–3 hours (Production). Sandbox returns instantly.
Sandbox returns Tesla, Inc. or BYD fixture data with the same structure as Production. 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": "corporate_exception_report",
"stage": "completed",
"progress": 100,
"content": {
"type": "results",
"data": {
"languages": ["en", "zk"],
"report_infos": {
"en": {
"report_date": "2025-11-30",
"report_name": "Corporate Exception Report — Tesla, Inc.",
"report_type": "enterprise changes report"
},
"zh": {
"report_date": "2025-11-30",
"report_name": "企业异常报告 — Tesla, Inc.",
"report_type": "enterprise changes report"
}
},
"chapter_infos": [
{
"en": {
"chapter_name": "Compliance",
"sub_sections": [
{
"section_title": "Regulatory Actions",
"sub_sections": [],
"section_texts": ["No material regulatory enforcement actions detected in the monitoring window."]
}
],
"chapter_texts": ["### Summary\nCompliance posture remains stable with no critical exceptions flagged."]
},
"zh": {
"chapter_name": "合规风险",
"sub_sections": [],
"chapter_texts": ["### 摘要\n监控周期内未发现重大合规异常信号。"]
}
},
{
"en": {
"chapter_name": "Reputation",
"sub_sections": [],
"chapter_texts": ["### Media Exposure\nNegative media sentiment remains within normal baseline levels."]
},
"zh": {
"chapter_name": "声誉风险",
"sub_sections": [],
"chapter_texts": ["### 媒体曝光\n负面舆情处于正常基线范围内。"]
}
}
]
}
}
},
"metadata": { "credits_used": 0 },
"errors": null
}
When the agent cannot unambiguously match a company from text, it returns WAITING_USER with a Markdown prompt in content (string). Continue with the same task_id.
Turn 1 — submit company name (optional chapter_name):
curl -X POST "https://agent.supplygraph.ai/api/v1/agents/corporate_exception_report/run" \
-H "Authorization: Bearer <YOUR_API_KEY>" \
-H "Content-Type: application/json" \
-d '{"text": "Tesla, Inc. United States", "chapter_name": "ALL", "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 entity (same task_id):
curl -X POST "https://agent.supplygraph.ai/api/v1/agents/corporate_exception_report/run" \
-H "Authorization: Bearer <YOUR_API_KEY>" \
-H "Content-Type: application/json" \
-d '{"text": "yes", "task_id": "<task-id>", "stream": false}'
The agent activates the subscription and begins baseline data preparation. Poll status every ~10 seconds until TASK_COMPLETED (initial build ~2–3 h in Production), then retrieve the report via results.
After the initial build completes, use results only to fetch the latest snapshot — no further status polling is required during the subscription.
See Agent API §8 for the general multi-turn pattern.
| Method | ID / Tool | Documentation |
|---|---|---|
| A2A | corporate_exception_report |
a2a.md |
| MCP | corporate_exception_report |
mcp.md |
| Agent API | corporate_exception_report |
agent-api.md |
Quick examples: A2A / MCP · Agent API
Common codes → Agent API §10. Supports mode=cancel for subscription cancellation. Returns WAITING_USER for company confirmation.
Maintainer: info@supplygraph.ai
License: Proprietary / Internal
© 2025–2026 SupplyGraph AI. All rights reserved.