Frustration hit hard one too many times. A former software engineer in Seattle ditched debugging code for swinging hammers. He launched his own general contracting firm, Kolmo Construction. Then he did something more ambitious. He built a free online AI system that translates the region’s tangled web of building and zoning regulations into plain answers.

The tool lives at kolmo.io/permits. It draws on 457 verified rules spanning 80 jurisdictions across King, Pierce and Snohomish counties. Each answer points straight back to the official city document and exact text. No vague summaries. No frantic searches through municipal PDFs that contradict one another from one suburb to the next.

Users type a question. The system replies with citations. Want to know if a lot allows two detached accessory dwelling units under Seattle rules? The answer appears in seconds. Need to check whether replacing a water heater requires a permit? Same process. Tree removal. Heat pump installation. Plumbing upgrades. The database covers them all.

Its creator, whose name appears in GeekWire’s reporting as Matta, spent years in tech before switching careers. He saw the problem up close. Homeowners spiral. Contractors burn hours. Projects stall. Local governments drown in incomplete applications and repeated correction cycles.

But this isn’t just another chatbot. Behind the clean interface sits Google’s Gemini model. It runs on retrieval-augmented generation. The technique pulls exact regulatory text into the prompt instead of trusting the large language model to recall or invent details. Answers come grounded. Hallucinations stay low. Every response includes source links.

Matta went further. He exposed the full database through the Model Context Protocol. Third-party AI assistants can now query the rules directly. Developers and other tools gain structured access. The data becomes portable across platforms.

His timing feels precise. Seattle has poured resources into similar efforts. In 2025 Mayor Bruce Harrell ordered pilots to cut housing and small-business permit delays. One effort with CivCheck tested pre-screening for residential applications. Results looked promising. Application completeness checks hit 87% accuracy. Design compliance reached 92%. Intake review times dropped roughly 50%. Correction cycles fell 35%. Those figures come from the city’s own Building Connections blog post.

The city budgeted $750,000 in 2026 for an AI permitting tool. Part of that covers integration with its existing Accela system. Officials want reviewers focused on complex judgments rather than chasing missing forms. Early pilots suggest the approach works. Yet city tools target official workflows. Matta’s project aims at the public. Homeowners. Small contractors. Anyone staring at a confusing zoning map.

Permit delays hurt. A recent Urbanist article from October 1 detailed another round of fee increases for building and land-use permits. Volumes remain depressed. Master-use permits sit 60% below 2019 levels. Smaller projects dominate. Costs keep rising. A four-unit townhouse that once carried $12,214 in fees now approaches $16,000 after two years of hikes. Developers feel the squeeze.

Matta’s database doesn’t submit applications. It doesn’t replace city staff. It simply tells users what the rules actually say. That transparency matters in a region where one suburb’s fence height requirement differs from the next by inches that trigger expensive redesigns.

Other cities experiment too. Honolulu cut average residential permit decision times from 73 days to 32.5 days after adopting AI screening, according to a September HousingWire report. Denver approved a $4.6 million contract with a similar platform. Even international efforts surface. Dubai plans to approve certain villa permits in minutes using AI that checks architectural drawings against building codes. A October 2 Scene Now story outlined the phased rollout.

Private companies race ahead. Startups train models on decades of plan review data. One provider, Avolve, expanded its AI agents in early October to cover pre-application guidance and change detection on resubmissions. The EIN Presswire release highlighted training on more than 20 years of agency records.

Matta chose a different path. Free access. Open database. Citations first. His background in software gave him an edge. He understood how to structure messy data. He knew retrieval techniques could anchor answers in truth. General contracting taught him which questions actually matter on job sites.

The result feels practical. A homeowner in Shoreline wonders about deck construction. A contractor in Tacoma needs clarity on electrical upgrades. Both receive jurisdiction-specific answers with links. No paywall blocks entry. No account required.

Challenges remain. Regulations change. Cities update codes. The database demands constant maintenance. Matta must verify new rules and retire outdated ones. Accuracy depends on clean source documents. Some municipalities publish clearer data than others.

Yet the model scales. Eighty jurisdictions already covered. More can be added. The MCP exposure invites integration. Future AI agents could pull from this database while handling full project planning. A virtual assistant might check zoning, suggest material lists, flag permit needs and even draft initial applications.

Seattle’s own data portal shows the volume. Nearly 200,000 building permits sit in open records. Many trigger multiple reviews. Backlogs persist despite good intentions. Tools like Matta’s attack the information gap that creates so many of those loops.

His project also highlights a broader shift. Technologists who once optimized cloud infrastructure now optimize government processes. They see bureaucracy as a data problem. They treat regulations as structured text that can be queried reliably. The approach won’t eliminate all disputes. Human reviewers still interpret edge cases. But it removes the guesswork that wastes everyone’s time.

Matta built Kolmo Construction first. The AI tool followed from lived experience. That sequence carries weight. He encountered the pain points as a contractor. He applied engineering discipline to solve them. The outcome serves both homeowners and fellow builders.

Look closer and the implications widen. If one former engineer can map three counties’ rules with citations, larger efforts could cover states or standardize certain building codes nationally. Federal grants now support AI adoption in local permitting. Interest grows.

For now the tool stands as a quiet success. Users ask. The system answers with sources. Projects move forward with fewer surprises. In a region desperate for more housing and faster construction, that progress counts. Simple. Direct. And long overdue.