AI Code Search

Written By Sarthak Kamra

Last updated About 1 month ago

What is AI Code Search?

AI Code Search is MeltPlan’s building code research and analysis agent that takes your code query, and provides reliable code requirements along with full explanations.

The agent considers your queries as either simple or complex.

For simple queries that require researching only a single code section, the agent instantly provides a reliable code requirement along with section citations.

For complex queries that require researching multiple code sections, exception handling, and other nuances, the agent goes into a deep research mode and provides a deeper code requirement along with full explanation of the entire interpretation path, along with personalized visual explaining your exact code requirement.

Why AI code research is now non-negotiable?

Checking for code compliance of designs can be extremely time consuming. Sometimes, it even takes days or weeks to solve a complex code scenario.

With Melt Code, you get an expert code research agent that understands it all, navigates codebooks for you, and instantly gives you the best possible answer with citations

What’s in this section?

How to use AI Code Search: A step-by-step guide

Step 1: Click ‘Code Search’ on the sidebar
Navigate to the left pane and select ‘Code Search’ to start a new query.

Step 2: Apply filters for better answers
Tell the AI where to look.

  1. Jurisdiction: Select the state, county, or city where your project is located so the AI searches the right set of adopted and amended codebooks for that area.

  2. Year: Choosing the correct year ensures that answers reflect the right version of the code – the one that actually applies to your project.

  3. Agency: By selecting the applicable agency (like OSHPD or DSA in California), you help the AI tailor results to the correct set of amendments and enforcement interpretations for your building type

Step 3: Type your query
Enter your code query, basic or complex. Just remember to provide context.

Step 4: Hit Enter
Within seconds, you'll receive a cited, summarized answer, tailored to your query.

In case of a simple query, the AI system provides an instant code requirement along with code section citations.

In case of a complex query, the AI system provides an instant code requirement along with code section citations.

Step 5: Answer clarifying questions
Adding details like occupancy or building type helps narrow results. Don't have them? Just click ‘Start Deep Research’.

Code Research for complex queries includes two views:

  1. Compliance Workflow

    A flowchart showing the step-by-step logic explanation the AI agent followed, with citations at each decision point.

  2. Compliance Scenario

    A personalized visual breakdown of requirements — like fixture counts, clearances, or spatial layouts — so you can see what compliance looks like in practice.

Step 6: View Compliance Workflow
See the full explanation (interpretation path that the agent followed). Each node shows a step in the logic with the relevant code citation cited.

Step 7: View Compliance Scenario
Switch to the Compliance Scenario tab to see a personalized visual of your exact compliance requirement - floor layouts, fixture placements, dimensional requirements, etc.

Pro Tips

  • Review both tabs: The workflow shows the entire logic explanation; the scenario shows a personalized visual. Together they give you the full picture.

  • Save to a project: Add the research results to a project to organize and collaborate with your team.

Next,

Step 8: Review actual code sections
Each answer comes with a direct citation – click the Source to jump to the exact code section it was pulled from. Easily verify and take code decisions with confidence.

Step 9: Save your research

After getting an answer, you can:

  • Save to Project: Add the query and answer to an existing project for future reference

  • Save to Checklist: Turn citations into checklist items—either all as one check or each citation separately

Quick Note

Queries you run in Code Search are private by default—only you can see them. Once you save a query to a project, it becomes visible to everyone in your organisation.

Step 10: Revisit your past queries
All your previous searches are saved in your personal history. You can go back anytime to reference past questions and answers without needing to retype or re-scope them.

Pro tips for writing query prompts for AI Code Search

MeltPlan AI system helps to research codes easily. When you ask a human a question, communicating the question properly helps you get a better answer.

Similarly, there are some pro tips to get the best answers from the AI. Below are a few of them to keep in mind:

💡Tip

🚫 Bad Query

✅ Good Query

Explanation

Not too long

In an existing 2 storey non-sprinklered type 5 building, what is the required fire separation between the ground floor B occupancy and the upper floor R3 occupancy?

In a non-sprinklered type 5 building, what is required fire separation between B occupancy floor and R3 occupancy floor?

Try to avoid any term that’s not helping in setting context for the question.

Eg. 2 storey, ground floor, and upper floor

Not too short

Fire separation B and R3

In a non-sprinklered type 5 building, what is required fire separation between B occupancy floor and R3 occupancy floor?

Try to add all terms that help add context to get the relevant answer.

Eg. non-sprinklered, type 5, B occupancy, R3 occupancy

Try changing query with keywords used in codes

What is the minimum width for a hallway in a hospital?

What is the maximum allowable length of a dead leg?

What is the minimum corridor width in a hospital?

What is the maximum allowable length of an uncirculated water pipe?

Codes use corridor vs. hallway

Dead legs might be used by people in trade lot, but code uses uncirculated water pipe

Powerful filters for AI Code Search

Get faster, more accurate answers by using filters that focus your search on what really matters.

Each filter helps Melt Code’s AI system zero in on the exact code sources relevant to your question.

Filters

Jurisdiction: Code interpretations vary widely across jurisdictions.

Select the state, county, or city where your project is located so the AI agent reseaches the right set of adopted and amended codebooks for that area.

Year: Every jurisdiction follows a specific code adoption cycle.

Choosing the correct year ensures that answers reflect the right version of the code – the one that actually applies to your project.

Agency: Different state agencies often adopt their own versions of building codes. By selecting the applicable agency (like OSHPD or DSA in California), you help the AI tailor results to the correct set of amendments and enforcement interpretations for your building type.

Benefits

  1. See the reasoning, not just the answer

    Understand how the AI arrived at its conclusion through a step-by-step workflow.

  2. Catch what you might miss

    Decision branches reveal exceptions, edge cases, and alternate paths you may not have thought to ask about.

  3. Visualize requirements

    Compliance scenarios show spatial layouts, fixture counts, and dimensional requirements in context.

  4. Export to documentation

    Use compliance workflows and scenarios in your code analysis reports.

Frequently Asked Questions (FAQs)

  1. Does every query go into the deep research mode?

    No. Deep Research activates for complex, multi-variable questions. Simple lookups return direct code requirements.

  2. Can I export the compliance workflow?

    Not currently. You can screenshot or save results to a project for reference.

  3. Can I add deep research results to a checklist?

    Yes. Click "Add to Checklist" to save citations from a Deep Research query. All citations gets saved as one check.

  4. Who can see my search queries?
    Queries are private until you save them to a project. Once saved, everyone in your organization can see them.

  5. What counts as a query?
    Each question you ask the AI counts as one query.

Happy building — and even happier code research!