Architecture Overview

🔧 Technical8 min readUpdated May 2026

System Overview

Gridlock is built as a multi-agent AI system where specialized agents work together to protect MSP client networks. Each agent is an independent unit with its own expertise, but they share intelligence and coordinate responses.

Agent Pipeline

When data flows in from the network, it goes through a processing pipeline:

1. Data Collection — The connector captures network traffic, system logs, endpoint data, and configuration snapshots
2. Normalization — Raw data is parsed into a standardized format the agents can understand
3. Routing — Data is routed to the appropriate agent based on type (network data → MSP Hunter, threat intel → Threat Researcher, etc.)
4. AI Processing — Each agent uses Z.AI GLM-5 to analyze data, identify patterns, and make decisions
5. Action — Agents take automated actions: block threats, generate tickets, update compliance scores, alert technicians

AI Provider

Gridlock uses Z.AI GLM-5 as its AI backbone. This gives us:

• Fast inference — Sub-second response times for threat detection
• Context awareness — Each agent maintains context about your specific environment
• Continuous learning — Agents improve their detection based on your network's patterns
• Cost efficiency — Enterprise-grade AI at a fraction of traditional SIEM costs

Data Flow & Security

Data flows through a secure pipeline:

1. Encrypted collection from endpoints via the connector
2. TLS-encrypted transmission to the Gridlock cloud
3. Processing in isolated agent environments
4. Results stored in encrypted databases
5. Dashboard access via authenticated sessions

Multi-Tenant Architecture

Each MSP gets a fully isolated environment. Your clients' data never touches another MSP's data. This is achieved through:

• Database-level tenant isolation
• Per-tenant encryption keys
• Scoped API access
• Separate agent contexts per client