AI Chat
The AI Chat panel lets you ask questions about your Blueprints, request explanations, and generate or modify graph logic — all within NodeLens, using your own API keys. Your data goes directly to the AI provider you choose, never through a third-party server.
AI Providers
NodeLens supports multiple AI providers. You can switch between them at any time from the dropdown in the chat footer. Each provider requires its own API key, which you enter once in Settings and is stored locally on your machine.
Claude (Anthropic)
Best overall performance for Blueprint analysis and code generation. Recommended for most users.
OpenAI (GPT)
Strong alternative with wide model selection. Uses the OpenAI API with your key.
Ollama
Runs models locally on your machine. No API key needed, no data leaves your PC. Performance depends on your hardware.
Custom / OpenAI-compatible
Any endpoint that follows the OpenAI API format — LM Studio, local servers, or custom deployments.
Loading Context
By default the AI has no knowledge of your Blueprints. You need to load context explicitly so it knows what you are working with. Click the + Context button in the viewer header to attach the current function to the chat.
There are two ways to load context: click the + Context button in the viewer header while a function is open, or right-click on a function in the sidebar and select + Context. You choose what to send — the content of the active viewer tab is attached as context. Each option has different trade-offs:
| Context type | Size | Best for |
|---|---|---|
| Pseudocode | Small | Understanding logic, asking questions about what a function does. Most token-efficient. |
| BSL | Small | Giving the AI a compact map of the graph with node IDs — useful before surgical edits. |
| BGL | Medium | When the AI needs to read and modify the graph. Full pin-level detail. |
| Graph JSON (Compact) | Medium | Raw graph data for advanced use — when BGL doesn't expose enough detail. |
| Graph JSON (Full) | Large | Complete data including all pins and defaults. Use only when necessary. |
Full Project Context via MCP
The in-app chat works on one function at a time. If you need the AI to understand the structure of your entire project — across multiple Blueprints and functions — the MCP Server is the right tool.
With MCP, an AI like Claude Desktop can read any function in your project, understand how they relate to each other, and make changes across multiple files in one conversation. It gives the AI a complete picture of your codebase rather than a single function snapshot.
See the MCP Server page for setup instructions.
Chat History
NodeLens automatically saves your chat history between sessions. The last 40 messages are stored locally and restored when you reopen the app — you can continue a conversation without losing context.
To save a conversation to a file, click the Export Chat button in the
chat panel. This saves the full conversation as a .txt file with a timestamp
in the filename (e.g. nodelens-chat-20260711.txt).
To start fresh, use the Clear Chat button to wipe the current conversation.
Custom Instructions
You can give the AI standing instructions that apply to every conversation — for example, preferred coding style, conventions specific to your project, or how you want answers formatted.
Go to Settings → Custom Instructions and enter your instructions in the text area. They are automatically appended to the system prompt at the start of every chat session.
Examples of useful custom instructions:
- "Always use Hungarian notation for variable names."
- "This project uses an inventory system based on grid coordinates. Struct FInventoryEntry has fields OccupiedCoord (FVector2D) and ItemClass (TSubclassOf)."
- "Keep explanations brief — I prefer code over prose."
- "When generating BGL, always confirm the pseudocode with me first before writing the graph."