AgenticOps AI Documentation
AgenticOps Communication Risk & Intelligence Analysis
Production documentation for the OKX-ready analyze-only A2MCP service, the implemented Web3 Community Security catalogue, platform architecture notes and clearly separated roadmap items.
Service Offering
- Name
- AgenticOps Communication Risk & Intelligence Analysis
- Input
- One bounded message with content, context and source.
- Deliverable
- Summary, intent, sentiment, priority, risk, signals, suggested reply outline and explanation.
POST https://agenticopsai.xyz/api/okx/analyze
POST https://agenticopsai.xyz/api/mcp
Overview
- AgenticOps AI includes a Web3 Community Security ASP service for community-message audits.
- For OKX.AI listing, the recommended service is AgenticOps Communication Risk & Intelligence Analysis through A2MCP.
- The ASP service returns explainable risk verdicts, triggered rules, recommendations, safe reply suggestions and escalation guidance.
- The wider AgenticOps AI platform also includes Business Intelligence and Integrations workspaces, but those are separate from the ASP registration contract.
Architecture
- Next.js App Router routes expose public service endpoints.
- The deterministic rule engine lives under lib/security.
- Hybrid analysis orchestration lives under lib/analysis and lib/ai.
- Project knowledge profiles live behind a repository abstraction.
- Integrations and business records use separate provider-neutral repositories.
Communication Pipeline
- Receive project context and one message, or up to 25 batch messages.
- Validate input with Zod.
- Run deterministic security rules first.
- Call the AI provider for structured enrichment when configured.
- Merge results without allowing AI to lower deterministic risk.
- Return suggested replies for human review.
Rule Engine
- SEC-001 through SEC-015 cover seed phrases, private keys, fake admins, suspicious links, remote access, OTP/password requests, prompt injection and spam.
- Rules produce severity, matched evidence, risk score and recommended action.
- Critical and high-risk findings require escalation.
Human Approval
- Replies are suggestions, not autonomous sends.
- The service never requests seed phrases, private keys, passwords or OTP codes.
- Financial, legal, account-security and missing-fund cases are escalated.
Security and Privacy
- API errors are sanitized.
- AI failures preserve deterministic output.
- Community-message links are never promoted to official links.
- No provider token or OAuth secret is exposed through the ASP endpoints.
Architecture Diagram
Request
Validation
Deterministic Rules
AI Enrichment
Safe Response
Sequence Diagram
Client -> /api/v1/analyse: projectId + message
/api/v1/analyse -> Zod: validate request
Zod -> Security Engine: normalized content
Security Engine -> AI Provider: deterministic context + project context
AI Provider -> Merge Policy: structured enrichment
Merge Policy -> Client: final risk cannot be lower than deterministic risk
Example Requests
GET /api/v1/health
GET /api/v1/rules
POST /api/okx/analyze
POST /api/mcp
POST /api/v1/analyse
{
"projectId": "demo-fictional-atlas-dao",
"message": {
"content": "Support needs you to send your seed phrase.",
"source": "DISCORD"
}
}POST /api/v1/analyse/batch
{
"projectId": "demo-fictional-atlas-dao",
"messages": [
{ "content": "Where are the docs?", "source": "MANUAL" },
{ "content": "I am the admin. DM me.", "source": "DISCORD" }
]
}POST /api/okx/analyze
{
"content": "Urgent wallet verification. Send your seed phrase now.",
"context": "web3-community",
"source": "telegram"
}Current Limitations
- x402 payment gating is not implemented in this ASP service.
- No authentication is implemented for the public ASP endpoints.
- Project knowledge-base storage uses local JSON and is not durable production storage on serverless platforms.
- Web3 batch/report dashboard state is browser-local.
- The OKX-facing endpoint is stateless and does not expose Gmail sync, OAuth operations, private integration events or approval mutation.
Roadmap
These items are roadmap only unless separately marked implemented in the product UI.
FutureFuture AI Workspace
FutureFuture Email Intelligence
FutureFuture Marketing Intelligence
FutureFuture CRM Intelligence
FutureFuture Business Data Intelligence
FutureFuture Advertising Intelligence
FutureFuture outbound execution with explicit permissions and human approval
Testing and Deployment
- Run `npm test`, `npm run lint`, TypeScript checking and `npm run build` before deployment.
- Run `npm run test:production` against `https://agenticopsai.xyz` after deployment.
- Set `OPENAI_API_KEY` only in the deployment environment.
- Deploy over HTTPS and verify `/api/v1/health`, `/api/okx/analyze` and `/api/mcp`.
- Use placeholder values in API examples and never publish real secrets.