Every software factory needs a maintenance layer.
MaintenanceOS is the layer that continuously maintains your software and infrastructure as the world around them changes, keeping you current, secure and compliant.
The build line
- Plan
- Design
- Build
- Review
- Test
- Ship
The maintenance layer · MaintenanceOS
A continuous loop over everything that ships
- 1 Identify Find what’s outdated or at risk
- 2 Understand Map impact, owners and dependencies
- 3 Prioritize Rank by risk, cost and urgency
- 4 Act Fix it with AI agents or your engineers
- 5 Verify Confirm the fix worked in production
- 6 Govern Keep every change within your policies
Continuous: every verified change resets the baseline for the next cycle.
Trusted by engineering teams at
Sound familiar?
You ship faster than ever. So does your maintenance backlog.
-
Tech debt
The backlog grows faster than you clear it.
Upgrades, deprecations and aging dependencies pile up behind feature work, until one of them becomes an outage.
-
End-of-life & lifecycle management
Operating systems, databases and runtimes reach end of support.
You find out weeks before end of support, and the upgrade needs prerequisites across teams nobody planned for.
-
PCI DSS & SOC 2 compliance
Audit findings trace back to unsupported software.
End-of-life runtimes, unrotated keys and configuration drift fail controls, and every audit turns into an engineering sprint.
-
Extended support costs
Paying to run past end of support.
Cloud providers and software vendors charge to keep end-of-life versions running. The fees hit the bill before anyone schedules the upgrade.
-
Vulnerabilities
CVEs in outdated images and dependencies.
Fixes wait in a queue because the component is old, the owner is unclear, and the upgrade path is unknown.
-
Certificates & secrets
TLS certificates move to 47-day lifetimes.
Manual renewals and key rotation don’t scale when every service has something expiring.
-
Ownership
Nobody knows who owns it.
What’s aging, who owns it and what happens next lives in spreadsheets, Jira backlogs and status meetings.
-
20-50%
of engineering capacity goes to reactive maintenance and tech debt management.
What changes
From catching up to staying ahead.
MaintenanceOS connects to the tools you already run and turns maintenance into a continuous loop your team and AI agents work through together.
Today
With MaintenanceOS
-
Tech debt tracked in spreadsheets and quarterly reviews
One live inventory of tech debt across cloud, code and runtimes
-
End-of-life upgrades planned the quarter they’re due
Every EOL date on one timeline, planned months ahead and sequenced across teams
-
Priorities set by whoever escalates loudest
Ranked by what breaks, what it costs, and which controls it violates
-
Maintenance tickets waiting for someone with time
Picked up automatically, with context and completion criteria
-
“Done” means the pull request merged
Done means verified in production, with evidence
-
An engineering sprint before every PCI or SOC 2 audit
Audit evidence attached to every change, continuously
-
Extended-support fees found on the cloud bill
Costs priced before they land, with the upgrade already planned
Wherever your software factory is today
Your software factory is a journey. Maintenance can’t wait for it.
Some teams run coding agents at scale. Most are somewhere in between. Either way, versions reach end of life, CVEs land and audits come around. MaintenanceOS works at every stage, and grows with you.
-
Engineers and tickets
Your engineers write the code. Maintenance lives in tickets, spreadsheets and quarterly reviews.
With MaintenanceOS
MaintenanceOS takes the recurring work off the backlog. Engineers approve changes and handle the exceptions.
-
AI-assisted engineering
Engineers ship faster with AI assistants. More code, the same team to maintain it.
With MaintenanceOS
MaintenanceOS keeps pace with what ships, and your assistants get Draftt’s context through MCP.
-
Agentic software factory
Coding agents build, test and ship around the clock. Change never stops.
With MaintenanceOS
MaintenanceOS gives your agents the right maintenance work, governs it, and verifies every change.
MaintenanceOS™
Same loop. Same policies. Same proof. At every stage.
If your teams already run coding agents, this is how the work divides:
| The job | General-purpose coding agents If your teams run them | MaintenanceOS
The maintenance layer |
|---|---|---|
| Find the work | Start from the task, ticket or prompt you assign. | Continuously detects maintenance obligations across cloud accounts, repos, registries and runtimes, ahead of each vendor’s lifecycle timeline. |
| Know the business context | Context from the repos and tools you connect. | Ownership, blast radius, cost of waiting and compliance scope for every finding. |
| Decide what matters first | Work in the order it’s assigned. | Prioritizes by impact and sequences prerequisites across teams and environments. |
| Write the change | Their core strength, including migrations and upgrades you assign. | Opens the pull request itself, or hands a scoped task to the agent you already run. |
| Govern execution | Within each tool’s own guardrails. | Your policies decide what runs autonomously; production stays gated on a human. |
| Prove it’s done | A merged pull request and passing CI. | Re-checks live production state, keeps the evidence, and reports cost per verified task. |
How it works
One loop for every maintenance workflow.
Operating system, database and runtime upgrades, dependency updates, certificate rotation, configuration drift. A Fortune 500 retailer used it to go from 20% to 90% of planned work done proactively.
1 Identify Next version · preview
See what changed, everywhere you run.
A version reaching end of support, a vulnerability, a new policy, a drifting config. MaintenanceOS connects to the tools you already run and turns every change into a maintenance obligation.
2 Understand
Know what runs, who owns it, and what it touches.
The Graph maps every obligation to the real resources behind it, the accountable owner, the consuming teams, the downstream services, and the cost of waiting.
3 Prioritize
Decide what gets fixed first, and why.
Every obligation is weighed by what it breaks, what it violates, what it costs, what it exposes and what it takes to fix, so the queue reflects business impact, not whoever escalated loudest.
4 Act
Hand the work to the right executor.
MaintenanceOS gives AI agents prioritized tasks, context, and completion criteria. It can route the same task to your coding agent, your workflow, or an engineer when judgment is needed.
5 Verify
Done means it worked in production.
A task only closes when MaintenanceOS re-checks live state and health against the completion criteria. Proof is kept with the change: what changed, what was done, and whether it worked.
6 Govern
Your policies decide. The record earns trust.
Approvals and policies set what agents may do alone. The record of what worked shows your teams where agents can take on more, per technology, per environment, even per change.
Fortune 500 retailer · MaintenanceOS™
30% less technical debt in six months, across 20 teams.
The platform team mapped 11 recurring maintenance workflows, started with two, and expanded across 20 teams and four divisions.
Work unit: verified production change Baseline: 20% proactive Period: 6 months
Proactiveness
90%
of planned work done proactively, up from 20%
Effort
95% less
~57,000 hours returned to other priorities
Delivery cost
$7.2M
lower per year; 80% lower cost per verified task
AI adoption
60%
7,200 of 12,000 tasks executed by agents
Completed work
12,000
verified production changes; 20 of 20 teams used the service
Technical debt
−30%
vs. an initial 10% goal in 60 days
The work with Draftt changed the mindset of the team. We made proactive maintenance a service capability: 12,000 tasks completed, 95% less demand on our engineers and $7.2 million lower annual delivery cost, with AI adoption across multiple divisions.
Customers
Hear it from them. Engineering leaders on governed maintenance.
“We were looking for a way to manage tech debt as a continuous discipline: mapped to an owner, planned into the roadmap, and addressed before deadlines became risks.”
Craik Pyke
Vice President, Infrastructure and Security Engineering, SurveyMonkey
ROI calculator
See what Draftt is worth to your engineering org.
Tell us your annual cloud spend. We’ll infer your environment complexity and estimate the annual value of continuous maintenance with Draftt.
Your environment complexity
- Engineering team
- 175
- Platform team
- 22
- EOL/EOS events, next 12 mo
- 39
Estimated annual exposure
-
Engineering capacity spent on maintenance
22 platform engineers × 4 hrs/week × $65/hr
$297K
-
Lifecycle events & extended support
39 events × $8K avg
$312K
-
Compliance & audit remediation
Reactive work on PCI, SOC2, ISO findings
$683K
Total annual tech debt exposure
$1.29M
Equivalent to losing 9.9 engineers of capacity per year
Draftt reclaims at least 50% of this exposure. That’s $646K back in your roadmap, or 5.0 engineers of capacity returned per year.
Get startedHow we calculate this
Methodology. Cloud Platform team estimated at 12.5% of engineering org (industry typical). Platform engineers spend 4 hours/week on maintenance program management (concrete time budget, midpoint of observed 2-8 hours/week range). Lifecycle events derived from cloud spend, assumed environment complexity, and Draftt’s lifecycle events database (~1.5 events per $1M cloud spend plus ~0.05 per engineer). Per-event cost based on published AWS extended support pricing ($0.60/cluster-hr EKS, $0.10/vCPU-hr RDS) plus engineering response time. Compliance work: 3% of total team time, based on PCI, SOC2, and ISO27001 audit cycles. Loaded engineering cost: $65/hr ($130K fully-loaded annual), reflecting blended cost across globally-distributed engineering teams (US, EU, LATAM, India). FTE equivalent: total exposure divided by fully-loaded annual engineer cost.
Use cases
One platform. Six ways teams put it to work.
Lifecycle management, compliance, tech debt, modernization, security and FinOps are applications of the same layer. Start with the one that hurts most today.
- Lifecycle Management Every end-of-life event, version drift and expiring certificate, on one timeline you can act on.
- Compliance The gaps that fail audits live in your infrastructure. Find them, fix them, and prove it, continuously.
- Tech Debt Every obligation mapped to an owner, prioritized by business impact, and tracked to verified closure.
- Modernization Upgrade plans, pull requests and agent hand-offs that move the stack forward on your schedule.
- Security Unsupported runtimes, unrotated keys and configuration drift, surfaced and governed by policy.
- FinOps Extended-support fees, idle capacity and waste, priced before they reach the cloud bill.
FAQ
Questions, answered.
What is Draftt?
Draftt builds MaintenanceOS, the maintenance layer for the software factory. It helps engineering teams find and fix technical debt, end-of-life upgrades, vulnerabilities and compliance gaps across their stack, continuously.
What is MaintenanceOS?
MaintenanceOS is Draftt’s platform. It connects to the tools you already run, identifies what changed, prioritizes what needs attention and coordinates your teams and AI agents to fix it under your policies and approvals, as a continuous loop: identify, understand, prioritize, act, verify, govern.
How does Draftt handle end-of-life (EOL) and end-of-support software?
It continuously detects operating systems, databases, runtimes, frameworks and managed services approaching or past end of support across your cloud accounts, repos, registries and runtimes. For each one it shows the affected resources and owner, sequences the upgrade plan and prices the cost of waiting, including extended-support fees.
How does Draftt reduce technical debt?
Draftt keeps a live inventory of technical debt, ranks it by what it breaks, what it costs and what it violates, routes each item to an accountable owner, and fixes it through upgrade plans, pull requests or AI agents, tracked to verified closure. A Fortune 500 retailer reduced technical debt by 30% in six months across 20 teams.
Do I need coding agents or an AI software factory to use Draftt?
No. Draftt works the same way whether your teams build with engineers, coding agents or both. If you already run coding agents, MaintenanceOS gives them the right maintenance work, with context and completion criteria, and checks the result.
Does Draftt change production systems on its own?
Only where you allow it. You choose the level of automation per technology and environment: an upgrade plan your engineers run, a pull request you merge, guided execution, or an agent-to-agent handoff. Production changes can stay gated on human approval.
What do I need to install?
Nothing on your hosts. There are no agents to deploy and no code changes required. Draftt connects through read-only API access to your cloud accounts, registries, source control and ticketing systems.
Which tools does Draftt integrate with?
AWS, Google Cloud, Azure, Kubernetes, MongoDB Atlas, Redis Cloud, JFrog Artifactory, Azure Databricks, OpenAI, AWS Bedrock, Azure AI Foundry, Tanium, Jira, ServiceNow, Azure DevOps, Slack, Microsoft Teams, email and Port.
How does Draftt help with PCI DSS and SOC 2 compliance?
It surfaces the infrastructure gaps that fail audits, such as end-of-life software, unpatched vulnerabilities and configuration drift, maps them to the controls auditors check, routes fixes to the owning teams and keeps an evidence trail for every change.
Is Draftt secure?
Draftt is SOC 2 Type II certified and GDPR compliant, uses read-only access and encrypts data at rest and in transit. Details are available in the Draftt Trust Center at trust.draftt.io.
Every software factory needs a maintenance layer.
Book a 30-minute call. Tell us where maintenance hurts today, and we’ll show you what MaintenanceOS would take off your team’s plate.