Enterprise Integration

CI/CD, SDK & Agent Telemetry

One API call to test any agent. Stream agent behaviors via SDK for continuous behavioral analysis. HMAC-signed webhooks and CI/CD gates enforce security at every release.

Partner API

REST APIbash

# 1. Authenticate

TOKEN=$(curl -s -X POST /api/v1/partner/auth \

-d '{"apiKey":"csk_live_your_key_here"}' | jq -r '.data.token')

# 2. Register agent (get agentId)

curl -X POST /api/v1/partner/agents/register \

-H "Authorization: Bearer $TOKEN" \

-d '{"name":"PaymentProcessor","endpoint":"https://..."}'

# → agentId: "ag_7f3a..."

# 3. Scan using agentId

curl -X POST /api/v1/partner/scan \

-H "Authorization: Bearer $TOKEN" \

-d '{"agentId":"ag_7f3a...","agentEndpoint":"https://..."}'

# 4. Query agent by ID

curl -H "Authorization: Bearer $TOKEN" \

/api/v1/partner/agents/by-id/agentId

# 5. Get scan report

curl -H "Authorization: Bearer $TOKEN" \

/api/v1/partner/scan/{scanId}/report

TypeScript SDKtypescript

import { MatrixShield } from '@matrixshield/sdk';

const cs = new MatrixShield({

apiKey: process.env.MATRIXSHIELD_KEY!

});

// Register agent → get agentId

const agent = await cs.registerAgent({

agentEndpoint: 'https://your-agent.com/invoke',

packageId: 'PS86efeLiItek9dSPPo1',

});

console.log(result.overallScore); // 82

console.log(result.grade); // 'B'

console.log(result.dimensionScores);

// CI/CD gate

const gate = await cs.gate({

agentEndpoint: process.env.AGENT_URL,

packageId: 'PS86efeLiItek9dSPPo1',

minGrade: 'B',

});

CI/CD Integration

GitHub Actions — PR Guard

on: [push, pull_request]

jobs:

security-scan:

runs-on: ubuntu-latest

steps:

- uses: matrixshield/scan-action@v1

with:

api-key: ${{ secrets.MATRIXSHIELD_KEY }}

agent-url: ${{ env.AGENT_URL }}

package-id: PS86efeLiItek9dSPPo1

min-grade: B # Block PR if grade < B

HMAC-Signed Webhooks

Scan results are delivered via HMAC-SHA256 signed webhooks. Verify the signature to ensure authenticity.

Header: X-MatrixShield-Signature

Payload: scan.completed, scan.failed

Retry: 6 attempts with exponential backoff

MCP Server Integration

Test agents exposed as MCP (Model Context Protocol) servers. MatrixShield connects as an MCP client via Streamable HTTP, discovers available tools, and runs security assessments through each tool endpoint.

+Register any MCP server URL — MatrixShield connects as a client

+Auto-discovers tools via tools/list and identifies chat endpoints

+Chat-compat mode: runs standard security packages through each tool

+Enterprise auth: bearer tokens, API keys, or custom headers

+Test all tools or select specific ones for targeted assessment

+Dynamic heuristic mode with schema-aware attacks — coming soon

MCP Registrationtypescript

// Register MCP agent

const agent = await ms.agents.register({

apiFormat: "mcp",

mcpUrl: "https://agent.com/mcp",

authMethod: "bearer",

authCredentials: "token...",

mcpToolName: "chat", // the one tool to scan

});

// Test connection (lists the tools the server exposes)

const test = await ms.agents.testConnection(agent.agentId);

// { tools: [{name: "chat"}, {name: "search"}], selectedTool: "chat" }

// Run scan against the registered tool

const scan = await ms.scans.create({ agentId: agent.agentId });

Agent Behavioral Telemetry

Go beyond point-in-time scanning. Integrate the MatrixShield SDK into your agent runtime to stream behavioral data continuously. Every tool call, every user interaction, every decision path is captured and analyzed for security anomalies.

Stream agent tool calls, responses, and decision paths to MatrixShield

Continuous behavioral analysis — detect drift, anomalies, and emerging risks

Agent inventory with ownership, risk tier, and environment classification

Alert policies trigger when behavior deviates from established baselines

Cross-reference behavioral data with adversarial test results

SDK Telemetrytypescript

import { MatrixShield } from '@matrixshield/sdk';

const monitor = new MatrixShield.Monitor({

apiKey: process.env.MATRIXSHIELD_KEY,

agentName: 'PaymentProcessor',

});

// Wrap your agent's tool calls

const result = await monitor.track({

action: 'tool_call',

tool: 'database_query',

input: userQuery,

output: agentResponse,

});

// MatrixShield analyzes behavior

// Dashboard shows real-time metrics

// Alerts fire on anomalies

Supported Agent Frameworks

LangGraph

Auto-detected

OpenAI Assistants

Auto-detected

Google Gemini

Auto-detected

Custom HTTP

Any endpoint

CrewAI

Via generic adapter

AutoGen

Via generic adapter

Amazon Bedrock

Via generic adapter

Anthropic Claude

Via generic adapter

CI/CD, SDK & Agent Telemetry

Partner API

CI/CD Integration

GitHub Actions — PR Guard

HMAC-Signed Webhooks

MCP Server Integration

Agent Behavioral Telemetry

Supported Agent Frameworks

Ready to Secure Your AI Agents?