We build AI into your business, and keep it running.

The team behind Komori will build your company's AI.

Inside Komori

Komori runs on the same systems we’d build for you.

Pick one to see how it works, and what it would do in your business.

A data pipeline is an assembly line for information. Documents go in one end; checked, organised facts come out the other, every day, with nobody copying and pasting.

Reads: Timely disclosures and annual reports
Runs: Every business day, as filings land
AI does: Extract, check, score
Produces: Scores, 10-year threads, signals

How it works

DataNew filingArrives from the exchange
AI modelReadAI pulls out the facts
AgentCheckAn agent re-checks hard cases
RulesScoreFixed rules score 12 aspects
ResultPublishFeeds the screener and alerts

A filing sentence turned into data example

We are raising our full-year operating profit forecast from ¥42.0bn to ¥48.0bn.

Type: Forecast revision
Metric: Operating profit
Before: ¥42.0bn
After: ¥48.0bn
Change: +14.3%

In your business

ContractsRenewal dates, obligations and risks, in one table
Support ticketsTopic, urgency and who should answer
Invoices and receiptsLine items ready for accounting

KomoriHone SystemsThe team that built Komori can build this for you.Book 30 minutes

Software breaks in quiet ways. Every part of Komori keeps a running record of what it's doing, and agents read that record around the clock, so problems are caught and explained before anyone notices.

Watched: Every service, job and AI call
Recorded: Logs, traces, metrics and cost
Responder: A triage agent
You get: A plain-language alert and a fix to review

How it works

DataSystems runServices, jobs and AI calls
DataRecordEverything they do is logged
RulesCheckHealth checks run on a schedule
AgentTriageAn agent reads the logs, finds the cause
ResultReportPosts to the team, opens a fix

What the team sees example

Triage agent · 06:14

Filing import is falling behind: 3 of 40 jobs failed since 06:10. Cause: the exchange changed its date format.

Fix proposed: accept both date formatsReady for review

Failed jobs per hour

In your business

Websites and appsOutages caught and explained early
Data feeds and integrationsBroken syncs flagged before reports go wrong
AI featuresCost and answer quality watched daily

KomoriHone SystemsThe team that built Komori can build this for you.Book 30 minutes

An agent is an AI that can use tools, not just talk. Ask it a question and it decides what to look up, reads the sources, and answers with links you can check.

Ask in: Plain Japanese or English
Tools: Search, filings, threads, prices
Answers: With a source for every claim
Works in: komori.app, Claude Code, Cursor

How it works

DataQuestionAsked in plain words
AgentPlanDecides what to look up
AI modelLook upCalls search and data tools
DataReadReads the actual filings
ResultAnswerEvery claim linked to its source

A question, the tool calls, the answer example

Which companies on my watchlist raised guidance this quarter?

Two: Hitachi raised operating profit from ¥42bn to ¥48bn, and Sony raised its dividend.

[1] Disclosure · 6501[2] Disclosure · 6758

In your business

Internal knowledgeStaff get answers from your documents, with sources
Sales and customersAn account brief in seconds before every call
OperationsAn agent that updates systems and files tickets

KomoriHone SystemsThe team that built Komori can build this for you.Book 30 minutes

AI can be confidently wrong. So before we trust it, we test it on real examples with known answers, and run those tests again every time anything changes.

Test cases: Real examples, answers checked by a person
Graded by: Fixed rules plus an AI grader
Runs: On every change
Shows: Score now vs. before

How it works

DataReal examplesWith answers a person checked
AgentRunThe system answers each one
AI modelGradeRules and an AI grader mark it
RulesCompareAgainst the last version
ResultShip or fixWorse scores don't go out

A test run, before and after a change example

BeforeNow

Facts extracted correctly94%96% ▲

Sources cited correctly98%98% ＝

Answer stays on topic91%87% ▼

One score dropped, so this change waits for a fix.

In your business

Before launchAgree together what a good answer is
After launchA weekly score you can read
New AI modelsProof the upgrade is actually better

KomoriHone SystemsThe team that built Komori can build this for you.Book 30 minutes

Companies rarely announce a change outright. The wording shifts a little each year. Komori lines up ten years of a company's filings and shows exactly how each topic moved, quoting the company's own words.

Reads: Up to ten years of filings per company
Finds: The same topic in every year
Marks: New, repeated, stronger, softer
Shows: The exact quote, dated

How it works

DataTen yearsEvery filing for the company
AI modelSplitAI breaks each into topics
AgentMatchLinks the same topic year to year
RulesMarkNew, stronger, softer, dropped
ResultThreadOne timeline per topic

One sentence over six years example

2019

“We rely on a single supplier for some materials.”

2022

“We rely on a single supplier for several key materials.”

2025

“We rely on a single supplier for our main materials.”

The same sentence got much stronger over six years.

In your business

Contract versionsWhat changed between this year's terms and last year's
Policies and manualsWhich rules moved, and when
Competitors' reportsHow their story is shifting, year by year

KomoriHone SystemsThe team that built Komori can build this for you.Book 30 minutes

When something happens, the obvious effect is rarely the whole story. Komori follows one event outward, step by step: who is hit first, who depends on them, and who is affected after that.

Starts from: One filing or news event
Follows: Up to five steps out
Uses: Supplier and customer links from filings
Gives: A tree you can open branch by branch

How it works

DataEventA filing or a news story
AI modelFirst effectsAI lists who is hit directly
AgentBranch outAn agent follows each link further
RulesPruneWeak links are cut
ResultMapEvery affected company, ranked

One event, followed outward example

EventThe yen strengthens sharply

1stExporters earn less per sale abroad

2ndTheir parts suppliers get smaller orders

3rdRegional banks lending to those suppliers feel it

1stImporters pay less for goods

2ndRetailers' margins widen

In your business

Supply chainIf one supplier stops, what else stops
Prices and ratesWhat a cost or rate change does to your business
Key customersEarly warning when a big customer is under pressure

KomoriHone SystemsThe team that built Komori can build this for you.Book 30 minutes

Nobody has time to read everything. Each morning Komori gathers what happened to the companies on your list, writes it up in a few short lines, and sends it before the market opens.

Covers: Every company on your watchlist
Reads: Yesterday's filings, news and prices
Arrives: 7:00, before the market opens
Length: A few lines, with links

How it works

DataCollectFilings, news, prices
RulesRankWhat matters most comes first
AI modelWriteAI writes a short summary
AgentCheckEvery number checked against the source
ResultSendIn your inbox at 7:00

This morning's email example

7:00 Morning recap · Semiconductors (12)

▲Tokyo Electron rose 4.1% after raising its full-year forecast.

▼Advantest fell 2.3%; no new filing, sector-wide selling.

●Disco files its annual report today. Watch the order backlog.

Written from today's 4 disclosures and 9 news stories

In your business

Sales and KPIsYesterday's numbers, explained in three lines
Customer supportWhat customers complained about this week
Market and competitorsA morning brief on what changed outside

KomoriHone SystemsThe team that built Komori can build this for you.Book 30 minutes

An idea sounds good until you check it. Komori replays a list of companies or a screening rule against past prices, so you can see what would actually have happened, good years and bad.

Tests: A watchlist or a screening rule
Against: Past prices and the TOPIX index
Reports: Return, worst drop, by period
Takes: Seconds

How it works

DataYour ideaA list or a rule
DataHistoryYears of daily prices
RulesReplayDay by day, as if you'd held it
RulesCompareAgainst the index
ResultResultReturn, drops, best and worst years

A list, replayed over twelve years example

Your listTOPIX

Return

+22%

Worst drop

-9%

Beat the index

+14pt

In your business

Pricing changesWhat last year's data says a price change would do
Rules and thresholdsTry a new approval rule on past cases first
ForecastsHow accurate a forecast method would have been

KomoriHone SystemsThe team that built Komori can build this for you.Book 30 minutes

We’re applied AI nerds.

We love figuring out how AI can solve a problem someone deals with every day. Give us a slow process or a tool you wish existed and we’ll work through it with you.

We build agents, applications, and the connections to your existing systems that make them useful.

ONE EXAMPLEA job from your inbox

An email comes in

A request, an address, a few missing details.

AI prepares the job

Find the details. Flag what needs checking.

Your team checks it

A work order, ready in the tool you use.

We’ve spent years building AI.

We’ve worked in AI, machine learning, and software engineering for years. We test systems against real examples, measure what they get right, and investigate what they miss.

You get software built for daily use, with a record of what happened and monitoring that helps us catch problems early.

Komori is built the same way. You’ve been using it.

THE ENGINEERING KITHow we check our work

Evaluations
Test on real examples.
Reliability
Keep a fallback ready.
Audit trails
Trace each decision.
Measurement
Track quality and cost.

We keep your AI up to date.

Hone Edge uses the map we keep of your workflows and systems. When a new model or tool appears, we check where it could help and test promising changes.

You get a short email with the likely benefit, cost, and our recommendation. You approve any change.

Reviews and proposals are included in ongoing support. Implementation is scoped separately.

Hone EdgeEXAMPLE EMAIL

TO Your teamFor your review

Testing a new model for email routing.

A decision model could handle your inbox categories. We’d test whether it can match your current system at a lower cost.

Incoming email

Decision modelPROPOSED CHANGE

SalesNew enquiry

AccountsInvoice

SupportHelp request

WHAT STAYS

Your inbox, categories, and routing rules.

WHAT CHANGES

The model that chooses the category.

We’ll compare accuracy and cost on your past emails, then send you the results and a price for the change.

You decide whether we go ahead.Hone

You save money and stay in charge.

We build tools that make the work easier and cheaper to do. We agree what to measure before we build, so you can judge the result.

You choose the priorities and approve the work. We handle the engineering and maintenance, and you keep the code, data, and infrastructure.

Start with a conversation. Discovery is free.

WHAT WE WORK TOWARDAgreed before we build

Less repetitive work

Give your team time back.

Lower running costs

Spend less on day-to-day use.

You stay in control

You set the priorities.

You keep the code, data, and infrastructure.

We build AI into your business, and keep it running.

The team behind Komori will build your company's AI.

Book 30 minutes

Inside Komori

Komori runs on the same systems we’d build for you.

Pick one to see how it works, and what it would do in your business.

A data pipeline is an assembly line for information. Documents go in one end; checked, organised facts come out the other, every day, with nobody copying and pasting.

Reads: Timely disclosures and annual reports
Runs: Every business day, as filings land
AI does: Extract, check, score
Produces: Scores, 10-year threads, signals

How it works

DataNew filingArrives from the exchange
AI modelReadAI pulls out the facts
AgentCheckAn agent re-checks hard cases
RulesScoreFixed rules score 12 aspects
ResultPublishFeeds the screener and alerts

A filing sentence turned into data example

We are raising our full-year operating profit forecast from ¥42.0bn to ¥48.0bn.

Type: Forecast revision
Metric: Operating profit
Before: ¥42.0bn
After: ¥48.0bn
Change: +14.3%

In your business

ContractsRenewal dates, obligations and risks, in one table
Support ticketsTopic, urgency and who should answer
Invoices and receiptsLine items ready for accounting

KomoriHone SystemsThe team that built Komori can build this for you.Book 30 minutes

Watched: Every service, job and AI call
Recorded: Logs, traces, metrics and cost
Responder: A triage agent
You get: A plain-language alert and a fix to review

How it works

DataSystems runServices, jobs and AI calls
DataRecordEverything they do is logged
RulesCheckHealth checks run on a schedule
AgentTriageAn agent reads the logs, finds the cause
ResultReportPosts to the team, opens a fix

What the team sees example

Triage agent · 06:14

Filing import is falling behind: 3 of 40 jobs failed since 06:10. Cause: the exchange changed its date format.

Fix proposed: accept both date formatsReady for review

Failed jobs per hour

In your business

Websites and appsOutages caught and explained early
Data feeds and integrationsBroken syncs flagged before reports go wrong
AI featuresCost and answer quality watched daily

KomoriHone SystemsThe team that built Komori can build this for you.Book 30 minutes

An agent is an AI that can use tools, not just talk. Ask it a question and it decides what to look up, reads the sources, and answers with links you can check.

Ask in: Plain Japanese or English
Tools: Search, filings, threads, prices
Answers: With a source for every claim
Works in: komori.app, Claude Code, Cursor

How it works

DataQuestionAsked in plain words
AgentPlanDecides what to look up
AI modelLook upCalls search and data tools
DataReadReads the actual filings
ResultAnswerEvery claim linked to its source

A question, the tool calls, the answer example

Which companies on my watchlist raised guidance this quarter?

Two: Hitachi raised operating profit from ¥42bn to ¥48bn, and Sony raised its dividend.

[1] Disclosure · 6501[2] Disclosure · 6758

In your business

Internal knowledgeStaff get answers from your documents, with sources
Sales and customersAn account brief in seconds before every call
OperationsAn agent that updates systems and files tickets

KomoriHone SystemsThe team that built Komori can build this for you.Book 30 minutes

AI can be confidently wrong. So before we trust it, we test it on real examples with known answers, and run those tests again every time anything changes.

Test cases: Real examples, answers checked by a person
Graded by: Fixed rules plus an AI grader
Runs: On every change
Shows: Score now vs. before

How it works

DataReal examplesWith answers a person checked
AgentRunThe system answers each one
AI modelGradeRules and an AI grader mark it
RulesCompareAgainst the last version
ResultShip or fixWorse scores don't go out

A test run, before and after a change example

BeforeNow

Facts extracted correctly94%96% ▲

Sources cited correctly98%98% ＝

Answer stays on topic91%87% ▼

One score dropped, so this change waits for a fix.

In your business

Before launchAgree together what a good answer is
After launchA weekly score you can read
New AI modelsProof the upgrade is actually better

KomoriHone SystemsThe team that built Komori can build this for you.Book 30 minutes

Reads: Up to ten years of filings per company
Finds: The same topic in every year
Marks: New, repeated, stronger, softer
Shows: The exact quote, dated

How it works

DataTen yearsEvery filing for the company
AI modelSplitAI breaks each into topics
AgentMatchLinks the same topic year to year
RulesMarkNew, stronger, softer, dropped
ResultThreadOne timeline per topic

One sentence over six years example

2019

“We rely on a single supplier for some materials.”

2022

“We rely on a single supplier for several key materials.”

2025

“We rely on a single supplier for our main materials.”

The same sentence got much stronger over six years.

In your business

Contract versionsWhat changed between this year's terms and last year's
Policies and manualsWhich rules moved, and when
Competitors' reportsHow their story is shifting, year by year

KomoriHone SystemsThe team that built Komori can build this for you.Book 30 minutes

When something happens, the obvious effect is rarely the whole story. Komori follows one event outward, step by step: who is hit first, who depends on them, and who is affected after that.

Starts from: One filing or news event
Follows: Up to five steps out
Uses: Supplier and customer links from filings
Gives: A tree you can open branch by branch

How it works

DataEventA filing or a news story
AI modelFirst effectsAI lists who is hit directly
AgentBranch outAn agent follows each link further
RulesPruneWeak links are cut
ResultMapEvery affected company, ranked

One event, followed outward example

EventThe yen strengthens sharply

1stExporters earn less per sale abroad

2ndTheir parts suppliers get smaller orders

3rdRegional banks lending to those suppliers feel it

1stImporters pay less for goods

2ndRetailers' margins widen

In your business

Supply chainIf one supplier stops, what else stops
Prices and ratesWhat a cost or rate change does to your business
Key customersEarly warning when a big customer is under pressure

KomoriHone SystemsThe team that built Komori can build this for you.Book 30 minutes

Nobody has time to read everything. Each morning Komori gathers what happened to the companies on your list, writes it up in a few short lines, and sends it before the market opens.

Covers: Every company on your watchlist
Reads: Yesterday's filings, news and prices
Arrives: 7:00, before the market opens
Length: A few lines, with links

How it works

DataCollectFilings, news, prices
RulesRankWhat matters most comes first
AI modelWriteAI writes a short summary
AgentCheckEvery number checked against the source
ResultSendIn your inbox at 7:00

This morning's email example

7:00 Morning recap · Semiconductors (12)

▲Tokyo Electron rose 4.1% after raising its full-year forecast.

▼Advantest fell 2.3%; no new filing, sector-wide selling.

●Disco files its annual report today. Watch the order backlog.

Written from today's 4 disclosures and 9 news stories

In your business

Sales and KPIsYesterday's numbers, explained in three lines
Customer supportWhat customers complained about this week
Market and competitorsA morning brief on what changed outside

KomoriHone SystemsThe team that built Komori can build this for you.Book 30 minutes

An idea sounds good until you check it. Komori replays a list of companies or a screening rule against past prices, so you can see what would actually have happened, good years and bad.

Tests: A watchlist or a screening rule
Against: Past prices and the TOPIX index
Reports: Return, worst drop, by period
Takes: Seconds

How it works

DataYour ideaA list or a rule
DataHistoryYears of daily prices
RulesReplayDay by day, as if you'd held it
RulesCompareAgainst the index
ResultResultReturn, drops, best and worst years

A list, replayed over twelve years example

Your listTOPIX

Return

+22%

Worst drop

-9%

Beat the index

+14pt

In your business

Pricing changesWhat last year's data says a price change would do
Rules and thresholdsTry a new approval rule on past cases first
ForecastsHow accurate a forecast method would have been

KomoriHone SystemsThe team that built Komori can build this for you.Book 30 minutes

We’re applied AI nerds.

We love figuring out how AI can solve a problem someone deals with every day. Give us a slow process or a tool you wish existed and we’ll work through it with you.

We build agents, applications, and the connections to your existing systems that make them useful.

ONE EXAMPLEA job from your inbox

An email comes in

A request, an address, a few missing details.

AI prepares the job

Find the details. Flag what needs checking.

Your team checks it

A work order, ready in the tool you use.

We’ve spent years building AI.

We’ve worked in AI, machine learning, and software engineering for years. We test systems against real examples, measure what they get right, and investigate what they miss.

You get software built for daily use, with a record of what happened and monitoring that helps us catch problems early.

Komori is built the same way. You’ve been using it.

THE ENGINEERING KITHow we check our work

Evaluations
Test on real examples.
Reliability
Keep a fallback ready.
Audit trails
Trace each decision.
Measurement
Track quality and cost.

We keep your AI up to date.

Hone Edge uses the map we keep of your workflows and systems. When a new model or tool appears, we check where it could help and test promising changes.

You get a short email with the likely benefit, cost, and our recommendation. You approve any change.

Reviews and proposals are included in ongoing support. Implementation is scoped separately.

Hone EdgeEXAMPLE EMAIL

TO Your teamFor your review

Testing a new model for email routing.

A decision model could handle your inbox categories. We’d test whether it can match your current system at a lower cost.

Incoming email

Decision modelPROPOSED CHANGE

SalesNew enquiry

AccountsInvoice

SupportHelp request

WHAT STAYS

Your inbox, categories, and routing rules.

WHAT CHANGES

The model that chooses the category.

We’ll compare accuracy and cost on your past emails, then send you the results and a price for the change.

You decide whether we go ahead.Hone

You save money and stay in charge.

We build tools that make the work easier and cheaper to do. We agree what to measure before we build, so you can judge the result.

You choose the priorities and approve the work. We handle the engineering and maintenance, and you keep the code, data, and infrastructure.

Start with a conversation. Discovery is free.

WHAT WE WORK TOWARDAgreed before we build

Less repetitive work

Give your team time back.

Lower running costs

Spend less on day-to-day use.

You stay in control

You set the priorities.

You keep the code, data, and infrastructure.

We build AI into your business, and keep it running.

Komori runs on the same systems we’d build for you.

We’re applied AI nerds.

We’ve spent years building AI.

Evaluations

Reliability

Audit trails

Measurement

We keep your AI up to date.

Testing a new model for email routing.

You save money and stay in charge.

Less repetitive work

Lower running costs

You stay in control

We build AI into your business, and keep it running.

Komori runs on the same systems we’d build for you.

We’re applied AI nerds.

We’ve spent years building AI.

Evaluations

Reliability

Audit trails

Measurement

We keep your AI up to date.

Testing a new model for email routing.

You save money and stay in charge.

Less repetitive work

Lower running costs

You stay in control