Real Estate AI Agent 'Tochi Kakaku Satei Kun' v1.4.2: The First Open-Source AVM Released Under Apache 2.0
Signal Yield Advisory has released 'Tochi Kakaku Satei Kun' v1.4.2, an open-source AVM embedded with the expertise of a veteran real estate appraiser, as a Claude Skill. Adopting a 'white box' approach that fully discloses its valuation logic, the tool is available on GitHub under the Apache 2.0 license. It aims to become the 'Linux of the AVM world' through community-driven development.
📋 Article Processing Timeline
- 📰 Published: May 21, 2026 at 19:00
- 🔍 Collected: May 21, 2026 at 10:31
- 🤖 AI Analyzed: May 21, 2026 at 11:02 (31 min after Collected)
Signal Yield Advisory (Roppongi, Minato-ku, Tokyo; Representative: Koichi Matsuda, Real Estate Appraiser) released the real estate AI agent 'Tochi Kakaku Satei Kun' (Land Price Assessor) v1.4.2 as a Claude Skill on May 21, 2026. This skill serves as an open-source AVM (Automated Valuation Model) designed to support land price brokerage assessments. It stands as the industry's first open-source AVM where a real estate appraiser with over 20 years of practical valuation experience, Koichi Matsuda, has published his comprehensive appraisal knowledge—including code, hedonic regression coefficients, and standard disparity table breakdowns—on GitHub under the Apache License 2.0 (based on company research as of May 2026).
Apache 2.0 is an industry-standard license equipped with patent protection clauses, belonging to the same lineage as TensorFlow and Kubernetes. The skill adopted this license specifically to lower legal barriers for corporate adoption. Utilizing the Ministry of Land, Infrastructure, Transport and Tourism's (MLIT) transaction price information CSVs and Land Price Publication GeoJSON (National Land Numerical Information L01) as input, the skill calculates valuation prices using a completely transparent 'white box' approach. It relies on hedonic log-linear OLS (Ordinary Least Squares) to regress coefficients such as area, walking distance from stations, and road widths on a per-property basis at the municipal level (meaning it is not a pre-trained model), fully disclosing even the breakdown of the disparity tables. Upon releasing the code, coefficients, and correction rates, 14 issues were simultaneously opened on GitHub Issues (including 3 Good-first-issues, 5 Appraiser review acceptance issues, and 1 Roadmap issue), officially launching its operation as a community-nurtured OSS.
This initiative is a proposal from a real estate appraiser driven by the ideal of becoming the 'Linux of the AVM world.' It can be installed into Claude Cowork / Claude Code with a single command and requires no usage fees (no additional costs beyond standard Claude subscription plans). Signal Yield Advisory is a PropTech firm with the mission of 'Building industry AI with our own hands.' This skill marks the second release in their Claude Skill series (following existing tools: Jyucho Kun and Kessai Kun).
## Background of Development: Aiming for the 'Linux of the AVM World'
We would like to clarify here why Koichi Matsuda, a real estate appraiser engaged in practical real estate valuation for over 20 years, decided to open-source all of his appraisal knowledge.
The core motivation behind releasing this skill was the desire to 'make this method easily accessible for anyone' when assessing land prices.
Just as Linux grew from an individual's decision to become the world's server infrastructure, this skill was released with the abstract metaphor of aiming to be the 'Linux of the AVM world.' It is a proposal from within the industry, provided as an OSS meant to be improved and nurtured by the community.
## Why Publish Under Apache License 2.0: An Industry Standard with Patent Protection
We adopted the Apache License 2.0 for this skill. Apache 2.0 is a widely used industry-standard license employed in massive OSS projects like TensorFlow (Google) and Kubernetes (CNCF).
Because it includes a patent protection clause, it prevents contributors and users from inadvertently facing patent litigation risks, thereby lowering the legal barriers for corporate business use, modification, and redistribution. Real estate appraisal offices, real estate brokers, and bank real estate departments can safely utilize and customize the tool for internal specifications without licensing concerns.
## 'An OSS Nurtured by the Community': Opening 14 Issues
Coinciding with the release of v1.4.2, 14 topics were raised in GitHub Issues.
- 2 Closed bugs: Published as a narrative demonstrating repository reliability.
- 3 Good-first-issues: Tasks requiring 30 minutes to 2 hours, welcoming engineers participating in OSS for the first time.
- 4 Intermediate issues: Improvement tasks aimed at mid-level developers.
- 5 Appraiser review acceptance issues (labeled 'needs-appraiser-review' + 'discussion'): Dedicated labels explicitly awaiting formal reviews from industry domain experts regarding topics such as regional stability verification of hedonic coefficients and validity discussions on the upper/lower limits of disparity correction rates.
- 1 Roadmap issue: Clarifying the future vision, with a planned transition to MLIT API integration in v2.0.
Gathering industry wisdom to nurture the OSS—this is the operational policy for this skill.
Issues URL: https://github.com/signal-yield/tochi-satei-kun/issues
Apache 2.0 is an industry-standard license equipped with patent protection clauses, belonging to the same lineage as TensorFlow and Kubernetes. The skill adopted this license specifically to lower legal barriers for corporate adoption. Utilizing the Ministry of Land, Infrastructure, Transport and Tourism's (MLIT) transaction price information CSVs and Land Price Publication GeoJSON (National Land Numerical Information L01) as input, the skill calculates valuation prices using a completely transparent 'white box' approach. It relies on hedonic log-linear OLS (Ordinary Least Squares) to regress coefficients such as area, walking distance from stations, and road widths on a per-property basis at the municipal level (meaning it is not a pre-trained model), fully disclosing even the breakdown of the disparity tables. Upon releasing the code, coefficients, and correction rates, 14 issues were simultaneously opened on GitHub Issues (including 3 Good-first-issues, 5 Appraiser review acceptance issues, and 1 Roadmap issue), officially launching its operation as a community-nurtured OSS.
This initiative is a proposal from a real estate appraiser driven by the ideal of becoming the 'Linux of the AVM world.' It can be installed into Claude Cowork / Claude Code with a single command and requires no usage fees (no additional costs beyond standard Claude subscription plans). Signal Yield Advisory is a PropTech firm with the mission of 'Building industry AI with our own hands.' This skill marks the second release in their Claude Skill series (following existing tools: Jyucho Kun and Kessai Kun).
## Background of Development: Aiming for the 'Linux of the AVM World'
We would like to clarify here why Koichi Matsuda, a real estate appraiser engaged in practical real estate valuation for over 20 years, decided to open-source all of his appraisal knowledge.
The core motivation behind releasing this skill was the desire to 'make this method easily accessible for anyone' when assessing land prices.
Just as Linux grew from an individual's decision to become the world's server infrastructure, this skill was released with the abstract metaphor of aiming to be the 'Linux of the AVM world.' It is a proposal from within the industry, provided as an OSS meant to be improved and nurtured by the community.
## Why Publish Under Apache License 2.0: An Industry Standard with Patent Protection
We adopted the Apache License 2.0 for this skill. Apache 2.0 is a widely used industry-standard license employed in massive OSS projects like TensorFlow (Google) and Kubernetes (CNCF).
Because it includes a patent protection clause, it prevents contributors and users from inadvertently facing patent litigation risks, thereby lowering the legal barriers for corporate business use, modification, and redistribution. Real estate appraisal offices, real estate brokers, and bank real estate departments can safely utilize and customize the tool for internal specifications without licensing concerns.
## 'An OSS Nurtured by the Community': Opening 14 Issues
Coinciding with the release of v1.4.2, 14 topics were raised in GitHub Issues.
- 2 Closed bugs: Published as a narrative demonstrating repository reliability.
- 3 Good-first-issues: Tasks requiring 30 minutes to 2 hours, welcoming engineers participating in OSS for the first time.
- 4 Intermediate issues: Improvement tasks aimed at mid-level developers.
- 5 Appraiser review acceptance issues (labeled 'needs-appraiser-review' + 'discussion'): Dedicated labels explicitly awaiting formal reviews from industry domain experts regarding topics such as regional stability verification of hedonic coefficients and validity discussions on the upper/lower limits of disparity correction rates.
- 1 Roadmap issue: Clarifying the future vision, with a planned transition to MLIT API integration in v2.0.
Gathering industry wisdom to nurture the OSS—this is the operational policy for this skill.
Issues URL: https://github.com/signal-yield/tochi-satei-kun/issues
FAQ
Is 'Tochi Kakaku Satei Kun' free to use?
Yes, there are no additional fees to use the tool, other than the standard Claude subscription fee.
Who developed 'Tochi Kakaku Satei Kun'?
It was developed by Koichi Matsuda (Representative of Signal Yield Advisory), a real estate appraiser with over 20 years of practical experience.
What does the 'white box' approach mean?
It means all calculation baselines, including source code, regression coefficients, and correction rates used for valuation, are fully transparent and disclosed.