Informat AI Agent
—— Automation Engine × AI Assistant, Building Enterprise-Grade AI Automation Platform
I. Why Choose "Informat AI Agent"
Against the backdrop of widespread attention on AI workflow platforms like Dify and Coze, the Informat AI Low-Code Platform doesn't simply "add an AI workflow capability"; instead, it deeply integrates AI into enterprise real business systems.
Unlike platforms that "only orchestrate AI Prompts and API calls", Informat AI Agent = Enterprise Business System × Automation Engine × AI Large Model × MCP Capability.
One-sentence Summary
Dify turns AI into workflows, Informat turns AI into truly productive "enterprise employees".
II. Informat AI Agent Overall Architecture
┌────────────────────────────────────┐
│ AI Agent │
│ ┌────────────┐ ┌──────────────┐ │
│ │ AI Assistant││ Automation Engine││
│ │ (Reasoning/Decision)││ (Process/Orchestration) │ │
│ └────────────┘ └──────────────┘ │
│ ↑ ↑ │
│ ┌────────────┐ ┌────────────┐ │
│ │ Data Tables/ │ │ MCP Server │ │
│ │ Business Systems││ External Capabilities │ │
│ └────────────┘ └────────────┘ │
└────────────────────────────────────┘
III. AI Assistant Capabilities (Intelligent Brain)
3.1 What is AI Assistant?
Informat AI Assistant is an AI Agent with enterprise context awareness capabilities, not a single-conversation robot, but:
- Understands your business model
- Remembers your data structure
- Continuously participates in business processes
- Can be called by automated processes at any time
3.2 Natively Understands Business Data (Core Advantage)
✅ Data Tables as Knowledge Bases (No Additional Modeling Required)
AI Assistant naturally understands business data tables in Informat:
- Table structure (field types, meanings)
- Table relationships (main table / sub-table / related table)
- Permissions and data isolation
- Real-time business data
Supported Operations:
- Query
- Insert
- Update
- Delete
IV. Automation Engine Capabilities (Execution System)
4.1 Automation ≠ Simple Workflow
Informat Automation is an enterprise-level process engine, not just "node connection".
Support:
- Conditional branching
- Parallel execution
- Loops
- Scheduled triggers
- Event triggers
- Exception capture and compensation
- Data table CRUD
- Workflows
- Dashboards
V. MCP Server: Connecting AI to the Real World
5.1 Informat Not Only "Calls MCP", But Can Also "Be MCP"
Informat AI Agent supports:
- As MCP Client
- As MCP Server
VI. Canvas Orchestration: What You See Is What You Get AI Automation
6.1 Unified Visual Canvas Capability
Informat AI Agent is based on a unified visual canvas, integrating the following capabilities in the same orchestration interface:
- 🤖 AI model calls
- 🔁 Automation process orchestration
- 🧠 AI reasoning and decision-making
- 🗂 Business data table operations
- 🔌 MCP Server tool calls
- 🌐 External API / Webhook
One-sentence Summary: On the same canvas, fully connect "thinking (AI)" and "execution (automation)".
6.2 Core Node Types Orchestrated in Canvas
1️⃣ AI Assistant Node
- Support for natural language instructions
- Support for structured output (JSON / Schema)
- Support for context memory
- Can read process context, business data, variables
Example Uses:
- Determine whether to enter the next process
- Generate business decision results
- Output parameters required for subsequent nodes
2️⃣ Automation Process Node
- Conditional judgment (If / Else)
- Parallel execution (Parallel)
- Loop processing (Loop)
- Scheduled trigger (Cron)
- Event trigger (data change / status change)
3️⃣ Data Table Node (Enterprise-Grade Core Capability)
- Query data
- Insert data
- Update data
- Delete data
- Support for permissions and tenant isolation
The biggest difference from platforms like Dify:
Operations here are on "real business databases", not temporary context variables.
4️⃣ MCP Tool Node
- Call external MCP Server
- Call third-party AI capabilities
- Call map / report / search / enterprise systems
6.3 A Complete Canvas Example (Text Version)
AI Customer Follow-up Agent:
[Scheduled Trigger] ↓ [Query CRM Customer Data] ↓ [AI Assistant: Analyze Customer Status] ↓ [Conditional Judgment] ├─ High Intent → [Generate Follow-up Suggestions] → [Notify Sales] └─ Low Intent → [Enter Nurturing Process]
The entire process:
- No code required
- All nodes are visual
- Full closed loop of AI decision-making + automated execution
6.4 From PoC to Production System
Informat Canvas is not a demo tool, but a business engine running directly in production systems:
- Support for high concurrency
- Support for permission control
- Support for logging and auditing
- Support for exception rollback and compensation
VII. Essential Differences Compared to Dify / Coze
7.1 Different Positioning Levels
| Platform | Core Positioning |
|---|---|
| Dify / Coze | AI workflow orchestration platform |
| Informat AI Agent | Enterprise-grade AI automation platform |
Dify solves "how to use AI",
Informat solves "how business runs".
7.2 Core Capability Comparison Table
| Comparison Dimension | Dify / Coze | Informat AI Agent |
|---|---|---|
| AI Workflow | ✅ | ✅ |
| Automation Capability | ⚠️ Basic | ✅ Enterprise-grade |
| Data Sources | Documents / API | Real business data tables |
| Permission System | Simple | Enterprise-grade RBAC / Multi-tenant |
| Complex Processes | Limited | Support for parallel / loop / events |
| MCP Capability | Client only | Client + Server |
| Secondary Development | Weak | Strong (Low-code + Extension) |
| Production Availability | PoC / Auxiliary | Core business systems |
7.3 Generational Gap in MCP Server Capability
Dify / Coze:
- AI can only "call external tools"
- Cannot be a capability provider
Informat AI Agent:
- AI can both call tools
- Can also provide business capabilities externally
For example:
Publish "customer query", "order creation", and "process approval"
Directly as MCP Server for external AI calls
7.4 Different Target Users
| Platform | Main Users |
|---|---|
| Dify / Coze | AI application developers |
| Informat AI Agent | Enterprises / Business systems / Middle platform teams |
7.5 Summary in One Sentence
Dify is an orchestration tool for AI,
Informat is an "onboarding system" for AI.
In Informat, AI has:
- Permissions
- Data
- Processes
- Responsibilities
- Results