RapidGPT v1.49 and Magnus v25.03: AutoPrompt Arrives - Multi-Message AI for Hardware Design

We're back with our March Release of 2025 for RapidGPT and Magnus (Enterprise Customers)!

note

For our RapidGPT users, RapidGPT v1.49 is now live on the VS Code marketplace! Updating to the latest version is seamless: if extension auto-update is enabled, the update will occur automatically when you reopen VS Code. For those who prefer manual updates, please refer to this guide on how to update an extension manually.

For Enterprise Customers, contact us today to learn more on how you can enable your hardware engineers with Magnus v25.03, the ultimate hardware AI assistant that understands your company's hardware data and documentation in a private setting.

Introducing AutoPrompt

Today, we’re excited to unveil AutoPrompt, our latest feature that breathes new life into your hardware engineering AI companion. AutoPrompt offers the generation of multiple AI messages per user query. By enabling AutoPrompt, RapidGPT/Magnus delivers a richer, more conversational experience—one that feels less like talking to a bot and more like collaborating with a trusted engineering colleague.

Traditional AI interactions can feel stiff: you pose a question, the AI responds with a single, often overwhelming chunk of information, and then the conversation ends. With AutoPrompt, RapidGPT/Magnus breaks down its responses into back-to-back messages—much like a human might do.

Built for Hardware Design

We built AutoPrompt specifically with hardware engineers in mind. Let’s say you’re designing a finite state machine (FSM). A single message might cover everything from high-level architecture to RTL code. That’s great, but maybe you want your AI to break down the process and "focus" on each aspect. With AutoPrompt, you can expect something like:

• Overview & Requirements – A quick rundown of the FSM’s goals, inputs, and outputs.
• State Breakdown – A message walking through each FSM state, its transitions, and what conditions trigger them.
• RTL Draft – A rough implementation example to show you how everything fits together.
• Follow-Up Q&A – A final check-in to address any open questions or concerns.

By spreading out the conversation, this allows RapidGPT/Magnus to "concentrate" on each individual step, while allowing you to digest each piece of information more effectively, identify gaps or mistakes, and immediately steer the discussion in a new direction if needed.

Why AutoPrompt Is Useful

• Human-Like Collaboration – The conversational flow helps you feel more connected to your AI partner—like having a seasoned engineer on your team 24/7.
• Clarity and Depth – Multiple, smaller messages make it easier to absorb complex hardware concepts step by step.
• Real-Time Guidance – You can jump in at any point with new ideas or questions, and the AI adjusts its approach without skipping a beat.

AutoPrompt: AI-Driven Hardware Design with Local PC Command Execution

With AutoPrompt, we’re exploring new ways AI can contribute directly to the engineering loop. Imagine your AI not only offering insights through a series of conversational messages but also taking ACTION to accelerate your workflow.

AI Terminal Control via Command Execution

By integrating with your local VS Code terminal, AutoPrompt brings your personal hardware AI assistant to a new level. Our previous release unlocked Command Execution, the ability for RapidGPT/Magnus to suggest terminal commands that the user may execute on the VS Code Terminal with a click of a button. We're excited to share that we've extended this terminal integration to AutoPrompt as well.

With proper command approval, RapidGPT/Magnus can autonomously EXECUTE terminal commands ON YOUR BEHALF.

Whitelisted Commands

Safety Warning

RapidGPT/Magnus in AutoPrompt Mode may suggest commands that will be automatically executed on your local machine. Only whitelisted commands in your RapidGPT VS Code settings are eligible to run, and you should always review and approve them before proceeding. While this helps maintain a safer environment, please remain vigilant—PrimisAI cannot be held responsible for any unintended effects. Use caution and confirm that each command is trustworthy before whitelisting.

Find the whitelisted command settings in RapidGPT settings in VS Code:

Settings -> RapidGPT -> Whitelisted Commands

A New Approach to Prompting AI for Hardware Engineering

With AutoPrompt + Command Execution, we're not just enabling better conversations—we're pioneering a new era of prompting, one that fuses AI reasoning with awareness of your local environment.

Think of it this way: your PC is no longer just the canvas for your ideas—it’s now an active tool in the AI’s toolkit. With AutoPrompt + Local Command Execution, your AI can see, understand, and interact with your project files, terminals, and toolchains in meaningful ways.

We're seeing a shift from static Q&A-style prompting to something interactive, contextual, and situationally aware. You’re no longer just asking for help—you’re delegating work to an AI that can take REAL action.

New Prompting Patterns We’re Unlocking

This new paradigm opens up entirely new ways of working with AI. You can now issue intuitive, task-oriented prompts that activate full workflows. For example:

• "Review my src/ folder and summarize the RTL architecture."
• "Check tb/uart_tx_tb.v and verify if the testbench stimulates all expected UART states."
• "Compile my design using iverilog with the -Wall flag and log warnings."
• "Run the testbench, extract the waveform, and tell me if the stop bit timing looks valid."
• "Compare power reports between synth_run1/ and synth_run2/, and summarize the PPA differences."
• "Generate a new state machine diagram from fsm_controller.v, and suggest optimizations."
• "Scan the Makefile for unsupported flags with our current simulator setup."

These aren’t just fancy prompts—they’re real tasks your AI can now handle by inspecting files, issuing safe commands, and reporting back with structured, intelligent responses. It's like having a junior engineer who not only understands hardware but also knows how to use your specific tools and file structures.

Demonstration: UART TX Testbench and Simulation with Icarus Verilog

To see AutoPrompt in action, check out our brief video demonstration where RapidGPT/Magnus creates a testbench for a UART TX module and runs a simulation using iverilog. You'll witness how RapidGPT/Magnus:

• Generates the testbench code for the UART Transmitter Module opened in the editor
• Executes terminal commands within VS Code to compile the design
• Runs the simulation to validate the functionality of the UART TX module.

The Perfect Trifecta: Leveraging Local Command Execution, AutoPrompt, and PAI Agents Together

Building on our previous release of PAI Agents, AutoPrompt empowers these specialist agents to operate autonomously with enhanced command execution capabilities. Imagine a suite of dedicated agents, each expertly tailored to a specific aspect of your design process:

• Project Investigator Agent: Your partner for deep project insight—scanning design files, testbenches, makefiles, TCL scripts, and documentation to build a comprehensive understanding of your work.
• Testbench Builder Agent: Your solution for robust testbench creation—iteratively generating and validating testbenches while seamlessly interfacing with local EDA tools.
• Simulation Analysis Agent: Your guide for simulation workflows—prepping simulations, operating EDA tools, analyzing outputs, and fine-tuning designs and testbenches for enhanced performance.
• PPA Optimizer Agent: Your expert for achieving optimal power, performance, and area—leveraging local synthesis tools and strategically tweaking TCL scripts and design parameters to drive PPA improvements.

With AutoPrompt and Local Command Execution, PAI agents are ready to elevate your workflow, becoming true specialist collaborators. Stay tuned for upcoming demos showcasing our agents in action!