AI Integration with MCP
NeoArc Studio includes an embedded MCP server that allows AI coding assistants to create documentation, manage data models, and query project structures directly.
AI-Assisted Documentation at Scale
NeoArc Studio embeds a Model Context Protocol (MCP) server directly inside the application. When enabled, any MCP-compatible AI tool can connect to your running instance and work with your documentation project using structured, validated operations.
This is not a generic AI integration. The MCP server understands NeoArc's file formats, content block types, data model structures, and validation rules. It guides AI assistants to produce correctly structured content and provides real-time feedback when something needs correcting.
How It Works
The MCP server runs as a local HTTP service inside the Electron application, listening on 127.0.0.1 only. When you enable it in Settings, AI tools can connect using the standard MCP protocol over HTTP. The server exposes tools that the AI can call, resources it can read, and instructions that guide its behaviour.
MCP Servers
NeoArc Studio exposes multiple logical MCP servers through a single connection endpoint. The AI tool sees all tools and resources from all servers and uses whichever is relevant to the user's request.
| Server | Tools | Purpose |
|---|---|---|
| Content | Tools and resources | Create pages, manage blocks, validate content, organise directories and sort order, bind content to publications |
| Model | Tools and resources | CRUD entities and properties, manage relationships, resolve foreign keys, impact analysis, governance auditing, model querying |
| Schemas | Tools and resources | Create API schemas, auto-generate from model with lineage, detect drift, check breaking changes, trace field lineage |
| REST API | Tools and resources | Create REST API projects, endpoints, security schemes (OAuth2, API keys, mTLS), servers, responses |
| Intent Graph | Tools and resources | Query the knowledge graph: impact analysis, lineage tracing, path finding, documentation coverage, orphan detection |
| GraphQL | Tools and resources | Create GraphQL APIs with queries, mutations, subscriptions, custom scalars, directives, interfaces |
| Async API | Tools and resources | Create event-driven APIs across a range of protocols (Kafka, AMQP, MQTT and others), channels, operations, messages |
| Webhooks | Tools and resources | Create webhook APIs with multiple verification methods, delivery/retry config, event types with payloads |
Compatible AI Tools
The MCP server works with any AI tool that supports the Model Context Protocol. This is an open standard with growing adoption across the industry.
Security and Deployment Options
The MCP server is designed with security as a priority. It is entirely optional, off by default, and can be fully removed from builds for environments that require it.
| Deployment Tier | MCP Available | Notes |
|---|---|---|
| Connected | Yes (off by default) | Full feature set. User enables in Settings when ready. |
| Internal | Yes (off by default) | Available for internal network use. Server binds to loopback only. |
| Air-gapped | No | MCP feature is completely removed. No server code runs, no port is opened, no settings are shown. |
| Custom | Configurable | Organisations can enable or disable MCP via the tier configuration file. |