Agentic AI for legacy modernization
Modernize the systems you can't afford to break.
Coddee.ai reads your entire legacy codebase, recovers the business logic buried inside it, and rebuilds it on a modern stack, one verified module at a time. It runs inside your environment, and your engineers approve every change.
C CONTNO CHAIN CONTRACT C IF %FOUND(CONTRACT) C EVAL BALANCE = PRINC C + LINKAGE + INTRST C ENDIF
def statement_balance(contract_no: str) -> Decimal | None: contract = contracts.find(contract_no) if contract is None: return None return (contract.principal + contract.linkage + contract.interest) # gate: output reconciles with the legacy system
Illustrative example
Why replacements fail
A replacement project fails when nobody knows exactly what the old system does.
Mapping a large system in full used to be measured in person-years, so in practice nobody did it. Teams sampled, asked around, and filled the gaps with assumptions. The gaps came back in year two, as overruns and wrong numbers.
The whole system, never a sample
Months of analyst work now takes days, so every program, table and interface gets read.
A specification that stays current
The spec is regenerated from the system itself, so it does not go stale while the project runs.
Gaps surface at the start
Open questions reach a business owner early, when answering them is cheap. That is the real saving.
What we do
One platform, three ways to put it to work.
Coddee.ai applies agentic AI to the engineering of large, long-lived software systems: understanding them, modernizing them, and keeping them maintained afterwards.
Legacy modernization
We move long-lived, business-critical systems onto modern architecture without losing the logic buried inside them.
How we modernize →Agentic AI engineering
A mesh of specialized AI agents analyzes, plans, codes, reviews and tests, inside your own network.
Explore the platform →AI training and adoption
Programs that take management and staff from curiosity to informed, practical use of AI at work.
See AI enablement →Before and after
Same rules. Same numbers. A different working day.
A buyer's account on a legacy AS/400 green screen, and the same account in the modernized application.
ACC110 BUYER ACCOUNT INQUIRY 06/10/26 DEMO01 14:32:07 Company . : 01 Site . : 0412 Bldg . : 03 Unit . . : 0022 Contract : 0048811 Buyer . . : DEMO BUYER A Ln Due date Cd Principal Linkage Interest Paid 01 15/07/25 01 250,000.00 4,120.50 0.00 254,120.50 02 15/01/26 01 400,000.00 9,870.25 0.00 409,870.25 03 15/07/26 02 400,000.00 7,410.00 612.40 0.00 04 15/01/27 02 350,000.00 5,902.75 0.00 0.00 05 DELIVERY 01 600,000.00 0.00 0.00 0.00 Balance due . . . : 1,363,925.15 Bottom F3=Exit F5=Refresh F9=Print statement F12=Cancel
Illustration with demo data. These are not actual customer screens.
How it works
Six agents. One engineering workflow.
Asking a single model to translate files gives you translated files. Coddee.ai runs a team of agents that mirrors an engineering team, and hands every result to your engineers for the final call.
- AnalyzerMaps modules, dependencies and interfaces
- PlannerOrders the work into dependency-aware tasks
- CoderImplements each task
- ReviewerChecks the result against the goals
- DebuggerValidates it with your build and tests
Customers
At work where mistakes are expensive.
A decades-old core system on IBM AS/400 runs sales, contracts, finance, guarantees and property registration for five departments. After mapping it end to end, we were awarded its full modernization onto a modern open-source stack.
Embedded software built over decades, in a codebase of millions of lines of C that can never leave the network. A fully on-premises proof of concept migrates representative subsystems to modern C++.
Who it is for
Built for regulated, high-stakes environments.
If these five sentences describe your organization, the platform was designed around your constraints.
Core systems that cannot simply be replaced
They accumulate decades of embedded rules, and the people who remember why retire faster than the code does.
Zero tolerance for financial error
Old and new run in parallel, and balances, guarantees and postings must reconcile before anything touches production.
Data that cannot leave the perimeter
Regulated data and proprietary code stay inside the network. The on-premises architecture was designed for exactly that.
Auditability as a baseline
Traceable changes, scoped permissions and human sign-off on every accepted modification are built into the workflow, not bolted on for the regulator.
A workforce that must be brought along
Adoption succeeds or fails on whether staff understand what the tools do and where the limits are.
AI changes the pace of the work, not the control of it.
Decisions and accountability stay with people. Every change the platform proposes is traceable and waits for an engineer to accept it. Nothing reaches production because an AI decided it should.
Start here
Bring us the system everyone is afraid to touch.
Start with a conversation. If there is a fit, we prove it on a small, representative part of your own code, inside your own environment.
The platform
A team of AI agents that works the way your engineers do.
Coddee.ai is an autonomous, asynchronous, multi-agent platform. It analyzes large codebases, converts legacy code to modern architecture, and turns tickets into reviewed pull requests while your team keeps working.
The agent mesh
Each agent has one job. The Orchestrator makes them a team.
- AnalyzerMaps modules, dependencies and interfaces
- PlannerOrders the work into dependency-aware tasks
- CoderImplements each task
- ReviewerChecks the result against the goals
- DebuggerValidates it with your build and tests
Analyzer
Builds a global picture of modules, dependencies and interfaces before a single line changes.
Planner
Turns the objective into ordered, dependency-aware tasks, so the work never becomes a blind file-by-file conversion.
Coder
Implements each task against the plan and the code around it.
Reviewer
Evaluates the result against the modernization goals and sends it back when it falls short.
Debugger
Validates the change through the build and test mechanisms you already have.
Orchestrator
Moves the work between the agents, looping back through coding and review as many times as the module requires.
Capabilities
Designed for code that cannot leave the building.
On-premises and secure
Source code, models, AI processing and generated output stay inside your controlled environment.
Asynchronous by design
Agents work in the background on long-running tasks. Nobody waits on a chat window.
Human in the loop
Every change arrives as a reviewable proposal. Your engineers accept, correct or reject it.
Continuous learning
Accepted transformations and reviewer corrections become reusable knowledge for the next module.
Fits your GitHub flow
Repositories in, pull requests out. Changes follow the review process you already run.
IDE-agnostic
No plugin to install and no editor to standardize on.
Day to day
From ticket to pull request while you work.
Outside a migration, the same agents clear routine development work in the background.
Receive a task
The agent picks up a ticket from the system you already use, such as Jira or Linear.
Resolve it on-prem
It analyzes the codebase, makes the change and validates it inside your environment.
Deliver a patch
The result lands as a GitHub pull request, ready for a developer to review.
Outputs
More than converted code.
Each run leaves behind the material a team needs to own the result.
Documentation
Generated from the code itself, so it describes what the system does today.
Architecture diagrams
Modules, data and interfaces, drawn from the real dependencies.
Tests
Created with the change and run before anything is offered for acceptance.
Security insights
Findings raised during analysis, ready for your team to triage.
Deployment
Runs where your security policy says it should.
On-premises
Fully inside your network, models included. The choice for organizations whose code can never be exposed to a public AI service.
Your cloud
Deployed in a cloud environment that you control.
Hybrid
Split between the two, following your security and regulatory requirements.
See it run on your own code.
Pick a small, representative module. We will show you what the agents make of it.
Solutions
From the first map to the last cutover, and the years after.
Start where your risk is. Some customers begin by finding out what their system really does. Others arrive with a migration already decided. Either way, the work moves in small, verified steps.
System discovery
Know exactly what your system does before you decide anything.
We read the entire system and write down what it does today: processes, business rules, data model, interfaces, and the open questions that need a business owner's decision.
- A knowledge book generated from the system itself
- A map of programs, data and interfaces
- A list of gaps and blocking questions, raised at the start
- A working mockup of the future system for your users to react to
Legacy modernization
Move the system without losing the logic inside it.
The hard part of an old system is never the screens. It is the logic that decides the numbers, written down nowhere except in the code. We reverse-engineer it and validate it against historical results.
- Strangler pattern: the legacy system keeps running until each replacement module is proven
- Old and new run side by side, and the results must reconcile before cutover
- The highest-risk logic becomes an independently testable service
- Delivery in module-sized increments, never a single replacement event
Program shape
What a full modernization covers.
A complete program runs in two phases, delivered in module-sized increments. For one large core system we were awarded in full, the first phase runs around 18 months.
Modernize the platform
The legacy system keeps running behind a stable interface until each replacement module is proven. Financial results run old and new side by side and must reconcile before cutover.
Add intelligence and keep it running
Once the platform is modern, its data becomes usable for analysis and forecasting, and the same agents keep the system maintained.
Continuous engineering
An on-prem agent that clears the backlog while your team builds.
After a modernization, or without one, the same platform takes routine tickets and turns them into pull requests your developers review.
- Repetitive development tasks are handled in the background
- Patches arrive through the review process you already run
- Every accepted fix makes the next one better informed
- Developers stay in charge: the agent assists, it does not replace
AI enablement
Bring your people along.
Training for management and staff, and an internal network of AI Business Partners that keeps adoption going after we leave.
- An AI literacy program for the whole organization
- A management briefing on AI and system replacement
- A five-stage journey to an AI excellence center
- Governance, toolkits and field mentoring that stay with you
Method
The method every engagement shares.
Map the whole system, not a sample
Understanding comes first, and it is generated from the system itself instead of assembled from interviews and assumptions.
Plan before generating
The work is broken into ordered, dependency-aware tasks. It is never attempted as a blind source-to-source conversion.
Move in verification-gated increments
Each module is coded, reviewed, tested and validated in an isolated environment before it is offered for acceptance.
Keep people in control
Every change is traceable and subject to human review. Nothing reaches production because an AI decided it should.
Run inside your walls
On-premises processing, isolated execution, scoped permissions and audit trails, with no need to expose proprietary code to public AI services.
Leave capability behind
Accepted transformations and reviewer corrections become reusable knowledge, and training builds the human side of the same capability.
Getting started
Every engagement starts small.
We open with a bounded proof of concept on real code, inside your own environment. No multi-year commitment up front.
You choose the code
A small but representative part of the existing system.
Access is scoped
Set up under agreed security conditions, limited to that selection only.
The system is mapped
Its structure and dependencies are analyzed first.
The agents run
Iterative coding, review and debugging on the selected modules.
Results are delivered
In a form your engineers can inspect, compare and validate.
We review together
A joint engineering review of correctness, architecture and code quality.
Whether the output is accurate and maintainable. Whether the platform understands dependencies or only translates files in isolation. Whether the workflow holds under your security requirements. Whether it scales beyond the selected modules.
Pick the module. We will bring the platform.
Tell us what the system does and what worries you about touching it.
Customers
Two very different organizations. One working method.
A residential developer and a defense electronics house have little in common, except core software that is decades old, impossible to pause, and too sensitive to hand to a public AI service. Together they cover both halves of what we deliver: deep modernization of a core system, and the program that brings an organization's own people up to speed on AI.
Modernizing the core system that five departments depend on.
This developer runs sales, contracts, finance, guarantees and property registration on a single core system written in RPG on an IBM AS/400. It serves five departments and connects to SAP, a CRM platform, business intelligence, banking interfaces and a buyer portal. Like most systems of its age, its full behavior lives in the code more than in any document.
Where we startedUsing the Coddee.ai platform, we carried out a full functional discovery and technical assessment: every process, business rule, data structure and interface, written up as a knowledge book and shown as a working mockup of the future application.
What we were awardedOn the strength of that work, the customer awarded Coddee.ai the entire modernization program: the full system, its data and its interfaces, moved onto a modern open-source stack in module-sized increments.
Where the difficulty sitsThe difficult part of a system like this is never the screens. It is the logic that decides the numbers and was never written down anywhere except in the code: the indexation and interest engine, discount ceilings, commission tiers, delivery-date recalculation, and the triggers that issue or release a statutory guarantee. That logic is reverse-engineered and validated against historical results, and it is built as an independently testable service because it carries the highest correctness risk in the project.
Where AI changes the economicsMapping a system this size in full used to be measured in person-years, so in practice nobody did it. We map the whole system instead of a sample, regenerate the specification from the system itself so it does not go stale, and surface the gaps at the start.
- Legacy
- RPG on IBM AS/400
- Target
- An open-source, containerized stack: Docker, PostgreSQL, Python services and a React front end
- Program
- Two phases. Phase A modernizes the platform across discovery, architecture, UX, security, data migration, integrations, build, testing and cutover. Phase B adds analytics, forecasting and ongoing operations.
- Timeline
- Phase A runs around 18 months, delivered in module-sized increments instead of a single replacement event.
- Approach
- Strangler pattern behind a stable interface. The legacy system keeps running until each replacement module is proven.
- Constraint
- Zero tolerance for financial-data error. Account statements, guarantees and accounting postings run old and new side by side and must reconcile to the cent before cutover.
Alongside the modernization, we run a structured AI education program for the same customer, because a modernized system in the hands of an organization that does not understand AI is a wasted investment.
- A two-part foundational course on what AI is: where it came from and how it is structured, then how it works
- A session on decision-making and human judgment in a world where software increasingly chooses for us
- A live podcast recorded in front of the audience, on AI in small and large companies and the dilemmas it creates for managers and staff
- A management briefing on why AI changes what a system-replacement project costs and how it can fail
Before and after
From green screen to an application people want to use.
Same data and the same business rules, shown the way a modern team expects to work. Three examples from a residential developer's core system.
ACC110 BUYER ACCOUNT INQUIRY 06/10/26 DEMO01 14:32:07 Company . : 01 Site . : 0412 Bldg . : 03 Unit . . : 0022 Contract : 0048811 Buyer . . : DEMO BUYER A Ln Due date Cd Principal Linkage Interest Paid 01 15/07/25 01 250,000.00 4,120.50 0.00 254,120.50 02 15/01/26 01 400,000.00 9,870.25 0.00 409,870.25 03 15/07/26 02 400,000.00 7,410.00 612.40 0.00 04 15/01/27 02 350,000.00 5,902.75 0.00 0.00 05 DELIVERY 01 600,000.00 0.00 0.00 0.00 Balance due . . . : 1,363,925.15 Bottom F3=Exit F5=Refresh F9=Print statement F12=Cancel
INV200 UNIT INVENTORY - WORK WITH 06/10/26 Site: 0412 DEMO PROJECT NORTH Bldg: 03 Opt Unit Fl Rm Area List price St _ 0021 05 4 112.0 2,640,000 S _ 0022 05 3 86.5 2,090,000 S _ 0023 06 5 131.0 3,120,000 F _ 0024 06 4 112.0 2,690,000 A _ 0025 07 4 112.0 2,740,000 A _ 0026 07 3 86.5 2,150,000 O _ 0027 08 5 131.0 3,260,000 S _ 0028 08 4 112.0 2,790,000 A More... 2=Change 5=Display 8=Price history F3=Exit F6=Add F12=Cancel St: A=Available S=Sold F=Frozen O=Owner
GUA300 BUYER GUARANTEES - WORK WITH 06/10/26 Site: 0412 DEMO PROJECT NORTH Opt Guar.no Unit Amount Bank Issued St _ 7700098 0019 245,000.00 12 04/11/24 R _ 7700121 0021 264,000.00 12 12/03/25 I _ 7700140 0027 326,000.00 31 03/02/26 I _ 7700163 0022 254,120.50 31 18/07/25 I _ 7700171 0022 409,870.25 31 20/01/26 I _ 7700188 0027 652,000.00 31 Q Bottom 5=Display 7=Release 9=Bank file F3=Exit F12=Cancel St: Q=Requested I=Issued R=Released
Illustration with demo data. These are not actual customer screens.
Modernizing embedded C that can never leave the network.
This organization maintains embedded software developed and extended over decades: large volumes of legacy C, hardware-specific logic, internal libraries, and architectural decisions layered across several generations of engineers. Manual modernization is possible but slow, and the code cannot be sent to any public AI service.
What we didA controlled, one-week proof of concept on two representative subsystems, migrating legacy C to modern C++ in incremental, verification-gated steps. The modernized modules keep a C-callable ABI, so they continue to work with the code around them.
How the work was performedNo single model was asked to translate files. A mesh of specialized agents mirrored the roles of an engineering team. An Analyzer built a global picture of modules, dependencies and interfaces. A Planner turned the objective into ordered, dependency-aware tasks. A Coder implemented them. A Reviewer evaluated the result against the modernization goals, and a Debugger validated it through the available build and test mechanisms. An Orchestrator moved the work between them, looping back through coding and review as many times as each module required.
Why it mattersThe point is not the line count. It is a repeatable, secure and auditable method for modernizing a very large body of legacy software, proven under strict security conditions.
- Codebase
- Millions of lines of legacy C
- Scope
- Two representative subsystems, C to modern C++
- Timeframe
- One week, structured to produce reviewable results instead of a demonstration.
- Deployment
- Fully on-premises. Source code, models, AI processing and generated output all stay inside the customer's controlled environment.
- Method
- Incremental, verification-gated migration steps, with C-callable ABI compatibility preserved.
- Control
- Every generated change is presented for engineering review. The customer's engineers accept, correct or reject it.
The platform does the heavy lifting. The customer's engineers keep engineering authority.
The common thread
What both engagements share.
The code stays inside
Processing happens where the customer's security policy allows, on-premises when required. Proprietary code is never exposed to a public AI service.
People sign off
Every change is traceable and subject to human review. Nothing reaches production because an AI decided it should.
Proof before scale
Each engagement began with a bounded piece of real work that could be inspected and judged. In real estate, that first phase led to the award of the full program.
Have a system like these?
Tell us about it. We will tell you plainly whether we can help.
AI enablement
A modern system needs people who know what to do with AI.
We run the training that takes an organization from curiosity to practical use, and we build the internal network that keeps adoption going after we leave.
Why it matters
Training alone does not change an organization.
Most organizations have exposed their people to AI. Far fewer have changed how the work gets done. Pilots follow the same shape, and the failure is almost never technological.
AI proofs of concept reach production in a typical enterprise.
Source: IDC research with Lenovoof the value in an AI transformation comes from people and process: workflow redesign, culture, governance and the way humans and AI work together.
Source: BCGThe causes are consistent, and human.
This is the gap our enablement work addresses, and it decides whether a modernization turns into measured value.
Fear of job loss
People hold back from tools they believe are meant to replace them.
No trust in the tools
Nobody explained what the tools do, or where their limits are.
Employees left out
Decisions about AI are made without the people whose work will change.
Shadow AI
People teach themselves outside work, and the organization learns nothing from it.
Track one
An AI literacy program for the whole organization.
Delivered alongside our modernization work at one of Israel's largest residential developers, for employees and management alike.
What is AI?
Part one covers where AI came from and how it is structured. Part two covers how it works, with a little maths.
Choosing in a world that chooses for us
Decision-making and human judgment when software increasingly decides for us, aimed at the practical and ethical questions employees already ask.
L & Y Talking AI
Recorded in front of the audience: AI in small and large companies, and the dilemmas it creates for managers and staff.
What AI changes in a system replacement
Twenty minutes plus questions on why AI changes what a replacement project costs and how it can fail.
AI changes the pace of the work, not the control of it. Decisions and accountability stay with people. An employee survey steers the content toward the topics people want covered.
Track two
An AI excellence center built on AI Business Partners.
An excellence center run only by IT produces solutions nobody asked for. One run by outside consultants produces knowledge that leaves when the retainer ends. We build a distributed network of people already inside your business units, in finance, legal, operations and compliance, who hold the domain knowledge, are willing to learn the tools, and are trusted enough by their peers to change how they work.
The organization
- Knows the unit's real information
- Characterizes the workflow
- Identifies the bottlenecks
- Understands the required outcome
The technology
- Commands the AI tool set
- Builds agents and automations
- Adapts tools to the unit
- Tracks what is newly possible
The people
- Trains peers directly
- Manages the change
- Makes tools and knowledge accessible
- Shifts mindset and motivation
Choosing the partners
Selection is the highest-leverage decision in the program.
It is not a matter of asking for volunteers. We look for five characteristics, and we assess them on evidence, not enthusiasm.
Knowledge foundations
Command of AI fundamentals and hands-on fluency across the working tool categories, beyond familiarity with one chatbot.
An AI-first mindset
The person already notices pain points in their own unit and proposes solutions, without waiting to be handed a use case.
A local pilot to their name
They have built something small that produced real value in a real workflow. This is the strongest single predictor.
Visible knowledge sharing
They learn continuously and pass it on, through one-to-one coaching, internal sessions or recorded material.
Standing to lead change
Close enough to the daily work to be credible, and trusted enough by peers that adoption follows them.
Adoption diagnostic
An organizational assessment that produces a maturity picture and a gap map for each unit.
Network analysis
Finds the people who are truly influential, who are not always the ones on the org chart.
Structured nomination
Run together with the department heads who will have to fund the time.
The common failure is naming AI champions and adding the role to their existing job. Below roughly a fifth of their time, the role dissolves into daily routine. The working standard is one to two protected days a week, written into performance objectives.
The journey
Five stages from mindset to practice.
AI mindset
The shift from executing tasks to designing and supervising agent systems, and the ground rules for trust, verification and data handling.
Business compass
Process analysis inside the unit: where work repeats, where decisions are rule-based, where errors and hours accumulate, and which of those are worth automating.
Development lab
Hands-on construction of the tools, agents and automations the unit needs, on the organization's own data and inside its security boundary.
Demo day
Solutions presented, success metrics defined before rollout, and a deployment plan agreed with the business owner.
Leading the change
Governance, information security, employee engagement and the knowledge-sharing routine that turns individual wins into organizational practice.
What stays with you
Capability that outlasts the program.
- A learning kit and an internal content library
- An adoption toolkit: a return-on-investment calculator for ranking candidate projects, characterization questionnaires and executive communication material
- A governance standard for policy, security and tooling, so that distributed building does not become uncontrolled building
- One-to-one field mentoring after the formal program ends, because the first two pilots are where these programs are usually lost
Automating a broken process
Streamline first, then automate. Accelerating an inefficient process only produces faster chaos.
No business ownership
Every initiative gets a named process owner from the business who is accountable for the result.
Optional change management
It is the most common cause of failure, and it needs a real share of the budget, not the remainder.
Weak governance
Uncoordinated initiatives create duplication and security exposure. A central center of excellence sets the standards the network works inside.
No measurement
Metrics are defined before the project starts. Otherwise value cannot be proven, and the budget does not survive its second year.
Plan an AI program your people will use.
Tell us where your organization is today, and we will suggest where to start.
Security
Your code stays inside your walls.
Coddee.ai was built for organizations that cannot send source code to a public AI service. The platform comes to the code, and the code stays where it is.
In the platform
Four controls built into every run.
On-premises processing
Source code, models, AI processing and generated output can all remain inside your controlled environment.
Isolated execution
Each module is built, tested and validated in an isolated environment before it is offered for acceptance.
Scoped permissions
Access is limited to the code selected for the work, under the security conditions agreed with you.
Audit trail and sign-off
Every change is traceable, and nothing reaches production without explicit human approval.
AI governance
The reviewing engineer owns the output.
Code changes, documentation, tests and recommendations produced with AI are reviewed by authorized engineers before acceptance. The engineer takes ownership exactly as if they had written it by hand.
Your code is not training data
Customer data and customer code are not used as general-purpose model training material by default.
Stricter handling on request
Closed-environment processing, code-only handling, data minimization and zero-data-retention configurations are available where you require them.
AI vendors are part of our supply chain
Any AI tool that touches customer information is evaluated for retention, training use, confidentiality and access controls.
In our own house
A small number of controls, applied without exception.
A startup cannot outspend an enterprise on security. It can choose high-leverage controls and apply them every day, with no carve-outs.
- Multi-factor authentication on every account, for everyone
- Least-privilege access, reviewed periodically
- Every code change peer reviewed through a pull request
- Separate development, staging and production environments
- Secrets kept in a vault, never in source code
- Data encrypted in transit and at rest
- Centralized logging and monitoring
- Access to a customer environment only with the customer's approval
Compliance position
Aligned, and clear about what that means.
Our practices are aligned with ISO/IEC 27001, the SOC 2 trust service principles, the OWASP Top 10, the NIST Cybersecurity Framework and the CIS Controls. We say aligned because that is what it is: alignment, not certification.
Where personal information is processed we apply privacy by design: data minimization, purpose limitation, restricted access and limited retention. Hosting region, residency and notification windows are agreed per engagement and written into the engagement documents.
Running a vendor security review?
Ask for our security overview. It separates what is implemented today from what is on the roadmap.
Company
A small team for very large systems.
Coddee.ai is an Israeli company building an enterprise AI platform for understanding, modernizing and accelerating complex software systems. We work with organizations whose software is old, critical and private.
What we believe
Four rules we do not bend.
AI moves the pace, not the control
Decisions and accountability stay with people. The platform proposes, engineers decide.
Understanding comes before code
We map the whole system before we change any of it.
Prove it small
Every engagement starts with a bounded proof of concept on real code, in the customer's own environment.
Say what is true
We separate what is done from what is planned, in our project plans and in our security documents.
Leadership
Leadership team.
A proven combination of operational excellence, technical innovation and strategic enterprise expertise.
15+ years in operations and customer management. Drives vision, go-to-market strategy, enterprise relationships and execution.
Operator and customer champion10+ years in machine learning and artificial intelligence. Architect of Coddee.ai's on-premises agent mesh and security design. Owns R&D, product development, security architecture and the technical roadmap.
Technical visionary30+ years in software engineering and enterprise sales. Provides strategic guidance on enterprise positioning, sales methodology and product-market fit.
Strategic advisor and industry veteranWhere we are
Based in Israel, working inside our customers' environments.
- Home
- Israel, with local support
- Team
- Remote-first engineering team
- Sectors
- Defense electronics, residential real estate
- Focus
- Legacy modernization, agentic AI engineering, AI training and adoption
- Contact
- contact@coddee.ai
Let's talk about your system.
A first conversation costs nothing and tells both sides whether there is a fit.
Talk to us
Tell us about the system.
We will tell you plainly whether we can help, and what a proof of concept on your own code would look like.
A first conversation
We listen to what the system does, how old it is, and what worries you about touching it.
A scoped proof of concept
You choose a small, representative part. We agree the security conditions together.
A joint engineering review
Your engineers inspect the results and decide whether they earn a next step.
Legal
Privacy Policy
What we collect through this website, why we collect it, and what you can do about it.
1. Who we are
Coddee.ai ("Coddee.ai", "we", "us") builds an AI platform for understanding and modernizing legacy software. This policy explains how we handle personal information collected through this website. For any question about it, write to contact@coddee.ai.
2. What this policy covers
This policy covers the Coddee.ai website and the messages you send us through it or by email. Work we carry out for customers, including any access to their code or data, is governed by the agreement and the security terms signed with each customer, and not by this policy.
3. Information we collect
- Details you send us. Your name, work email, company, the topic you choose, your legacy stack and your message, when you use the contact form or write to us by email.
- Technical information. Like most websites, the systems that host this site may record standard request data such as IP address, browser type, pages requested and time of visit.
- Cookies. This website uses only the cookies it needs in order to work. If we add analytics or marketing cookies in future, we will update this policy first and ask for your consent where the law requires it.
4. How we use it
- To reply to your request and continue the conversation you started.
- To assess whether and how we can help your organization.
- To operate, secure and improve this website.
- To meet our legal obligations.
We do not sell personal information, and we do not use the messages you send us for advertising.
5. Who we share it with
- Service providers that host this website and run our email and business tools, under confidentiality and security commitments.
- Public authorities, where the law requires it.
- A successor company, if Coddee.ai is involved in a merger, acquisition or sale of assets.
Some of these providers may process information outside your country. Where that happens, we rely on the safeguards that the applicable law requires.
6. How long we keep it
We keep your message and contact details for as long as we need them to handle your request and any business relationship that follows. If nothing follows, we delete them within a reasonable period. You can ask us to delete them sooner.
7. How we protect it
Access is limited to the people who need it, our accounts are protected with multi-factor authentication, and data is encrypted in transit and at rest. No system is perfectly secure, so please do not send source code, credentials or confidential business information through the contact form.
8. Your rights
Depending on where you live, you may have the right to see the personal information we hold about you, to correct it, to delete it, or to object to how we use it. This includes rights under Israel's Protection of Privacy Law and, where it applies, the EU General Data Protection Regulation. To use any of these rights, write to contact@coddee.ai.
9. Children
This website is intended for business users. It is not directed at children, and we do not knowingly collect their information.
10. Changes to this policy
We may update this policy from time to time. The date at the top of the page shows when it was last changed.
Legal
Terms of Use
The rules for using this website. Customer engagements are governed by their own signed agreements.
1. About these terms
These terms apply to your use of the Coddee.ai website. By using the site you agree to them. If you do not agree, please do not use the site.
2. What this site is for
The site describes Coddee.ai's platform and services. Its content is general information. It is not an offer, a quotation or a commitment to deliver. Any engagement with Coddee.ai is governed only by a written agreement signed by both sides.
3. Acceptable use
- Do not use the site for any unlawful purpose.
- Do not try to disrupt it, probe it for vulnerabilities, or gain unauthorized access to it.
- Do not use automated tools to copy substantial parts of its content.
- Do not misrepresent who you are when you contact us.
4. Intellectual property
The content of this site, including its text, graphics, design and the Coddee.ai name and logo, belongs to Coddee.ai or its licensors. You may view it and share links to it. You may not copy, modify or reuse it for commercial purposes without our written permission. The names and trademarks of other companies mentioned on the site belong to their owners.
5. Customer examples and code samples
Descriptions of customer engagements are summaries published for illustration. Results in one engagement do not guarantee results in another. Code shown on this site is illustrative and is not taken from any customer system.
6. Information you send us
Please do not send confidential information, source code or credentials through the contact form. Until a confidentiality agreement is in place, we cannot treat unsolicited material as confidential. How we handle personal information is described in our Privacy Policy.
7. Links to other sites
The site may link to websites we do not control. We are not responsible for their content or their privacy practices.
8. No warranty
The site is provided as is. We work to keep it accurate and available, but we do not promise that it is complete, current or free of errors.
9. Limitation of liability
To the extent the law allows, Coddee.ai is not liable for indirect or consequential loss arising from your use of the site or your reliance on its content. Nothing in these terms limits liability that cannot be limited by law.
10. Changes to these terms
We may update these terms from time to time. The date at the top of the page shows when they were last changed. Continuing to use the site after a change means you accept the updated terms.
11. Governing law
These terms are governed by the laws of the State of Israel. The competent courts of Tel Aviv-Jaffa have exclusive jurisdiction over any dispute relating to them.