Richtfest rebuilds construction planning as an AI-native process. Our proprietary agent system pairs LLM reasoning with fast System 1 models to test thousands of options per site, and finds the cost savings a three-person team never has time to look for.
External sound-insulation assessment, our first deterministic calculation workflow
A three-storey apartment house in Berlin, exactly as a first design might come in. Our agent system goes through it the way a thousand planners would, and shows you what it found.
One continuous pipeline from first site visit to execution drawings. People lead the project, agents do the legwork, and versioned tools do the maths.
Feasibility, massing, building physics and cost live in one shared project workspace, so a change on one sheet ripples through every other.
Planning fees are a small share of a project. Construction cost is the big one, and that is where a thousand minds pay for themselves.
Placeholder figures for layout. Replace with measured project data.
Developers, architects, contractors and authorities work from the same records, each with the view and permissions their role needs.
Nothing consequential leaves Richtfest without a person signing it off. Source files and revisions are kept intact, and every value carries its origin.
Our proprietary agent system combines two kinds of thinking. System 1 models judge thousands of options in milliseconds. LLM agents reason through the few that matter. Deterministic tools then check the result.
Fast, trained intuition. Purpose-built models trained on plans, quantities and cost data score a layout the way an experienced planner glances at it: instantly. They sweep through every grid, core position and facade ratio and keep only the promising ones.
Slow, deliberate reasoning. Specialised agents read briefs, zoning plans and stakeholder files, weigh trade-offs, explain their choices and coordinate the work. They decide which ideas the System 1 models should chase next.
One example project, 42 apartments on an urban infill plot. Each step is an optimisation the agent system found, checked and handed to a planner for approval.
Example project with placeholder figures. Replace with measured project data.
Send us a site and a brief. We will come back with your first variants.
