"Generate a demo sales dataset, write it into the database, and plot a sales trend chart."
During our recent webinar, DolphinX completed this entire workflow—from a single natural language request to a finished report in about one minute. Without writing Dlang code or performing manual operations, users simply described what they wanted, and the Agent handled data generation, database operations, and visualization automatically.

Enterprise AI is no longer limited by model capability. Today's challenge is deployment—connecting AI to enterprise data, workflows, and governance so it can operate safely in production.

On July 15, we hosted a webinar introducing DolphinX, our enterprise AI Agent development and governance platform. , the session focused on the practical challenges of deploying AI Agents in production, including permissions, security, governance, and enterprise integration.

Why Is Enterprise AI Still Hard?

Most organizations run into the same three walls:

• Limited access to enterprise data. General-purpose models need access to databases, sensors, business rules, and domain knowledge. Without it, they can only provide generic answers instead of solving real business problems.
• Security and permission concerns. AI systems often lack fine-grained access control, making organizations reluctant to allow them to perform production operations.
• Repeated development. Different teams frequently build similar Agents independently, leading to duplicated effort and higher maintenance costs.

Open-source frameworks are flexible but lack enterprise features such as governance, auditing, and permission management. Consumer AI platforms are easy to use b large-scale data processing and time-series analytics. Integrated commercial platfor b often difficult to deploy privately or adapt local regulatory requirements.

What enterprises need isn't another foundation model—they need an infrastructure layer that connects AI models with enterprise data, business capabilities, and governance.

That's exactly what DolphinX is built for.

Introducing DolphinX: An Enterprise-Grade Platform for Agent Development and Governance

Unlike standalone Agent frameworks, DolphinX is built directly into the DolphinDB Server. Through its built-in Coding Agent, users can interact with DolphinDB using natural language to perform:

• Data queries
• Script generation
• Batch processing
• Stream processing
• Data visualization

DolphinX sits in the platform layer of the DolphinDB AI architecture, combining DolphinDB's storage, computing, and AI capabilities with enterprise governance — a common foundation for industry-specific Agents.

It offers three deployment options:

• End users can use built-in Agents directly through the DolphinDB web interface.
• Business teams can visually configure prompts, Skills, MCP tools, and memory to build customized Agents without coding.
• Developers can integrate DolphinX APIs into their own applications for more advanced scenarios.

We've already shipped several production-ready Agents on top of DolphinX, including a FICC pricing agent and a strategy backtesting agent, plus reusable Skills for multiple tasks. More industry-specific Agents and Skills are on the way.

Four Core Capabilitie

Intelligent Context and Memory

Built-in intelligent context management and long-term memory enable Agents to retain context, learn from interactions, and better understand enterprise workflows over time.

Skills and MCP Integration

An extensible AI toolkit includes more 20 built-in Skills for common tasks for Dlang programming, data import, strategy backtesting, and machine learning. It also supports custom Skills and seamlessly integrates with internal and external tools and data sources through the standard MCP.

Secure Execution and Permission Management

Built on DolphinDB's mature permission system, DolphinX provides fine-grained control over data access and tool execution. Every AI operation is logged and fully auditable. Generated scripts are automatically inspected, and high-risk operations are blocked by default.

Enterprise Agent Governance

DolphinX provides a centralized management console for configuring Agents, models, Prompts, Skills, and MCP services. It also offers usage monitoring, cost analysis, and Skill version management to support enterprise-grade Agent operations.

Looking Ahead

DolphinX will continue evolving in four key areas:

• Multi-Agent collaboration for more complex workflows.
• Real-time AI by combining Agents with DolphinDB's streaming computing capabilities.
• Industry-specific ecosystems, with more Skills and tools for finance, manufacturing, energy, and other sectors.
• Enterprise governance, including standardized Agent evaluation and large-scale management.

Webinar Q&A Highlights

Q1: Can non-technical users perform data analysis using only natural language?

Yes. Enterprise users are already using DolphinX in production. The built-in CodingAgent can automatically complete data querying, analysis, visualization, and machine learning workflows without requiring Dlang programming.

Q2: Does DolphinX require an internet connection? Can it support offline document retrieval?

No. DolphinX can run entirely offline when deployed with a locally hosted LLM. Internet access is required only when using external LLM APIs, making it suitable for private deployments and air-gapped environments.

For offline knowledge retrieval, users can either build a custom document retrieval Skill or deploy an MCP-based document service within the private network to provide access to internal knowledge bases.

Q3: How does DolphinX handle permissions and security?

DolphinX inherits DolphinDB's native permission system. Access to data, databases, Skills, and MCP services is controlled according to user identity.

Different users receive different results based on their permissions, and all operations are fully audited. High-risk scripts are automatically blocked or can be configured for manual review before execution.

Q4: Are there production use cases? Does it support strategy backtesting and equipment fault prediction?

Yes. Current deployments include:

Financial Services: DolphinX has been deployed for FICC pricing and stock strategy backtesting. It supports natural language strategy creation and validates configurations before execution.

Power and Industrial Applications: AI Agents have been deployed for time-series analysis, equipment fault prediction, and load forecasting, automatically generating analytical reports from operational data.

Q5: Does DolphinX support long-term memory? Can it isolate multiple users and departments?

Yes. Conversation history is preserved as short-term memory. After periods of inactivity, useful business knowledge can be summarized into long-term memory.

Memory is separated into shared agent memory and user-specific private memory. Sessions remain isolated across users, while administrators can control which Agents, models, Skills, and tools are available to different departments.

Conclusion

The gap between "powerful model" and "AI running safely in production" is where most enterprise AI projects stall.

DolphinX is our attempt to close that gap — combining DolphinDB's data platform and compute engine with the governance layer enterprise deployments actually need.

Download the latest DolphinDB release and try DolphinX.