Software maintenance contract in Nepal: what an AMC covers
What a software maintenance contract (AMC) should cover, from severity levels to source code on exit, using our own AMC standard as the worked example.
EK Ra Sunya team
EK Ra Sunya Inc.
A software maintenance contract, usually called an AMC in Nepal, should pay for keeping a system you already own running, secure and current. It should not be a vague retainer for new features. Before you sign, check five things: what counts as maintenance, how fast each kind of problem gets answered, when support is open, where small changes end and paid changes begin, and what you walk away with if you leave.
A contract that is silent on any of them leaves the answer to whoever is angrier on the day. Below is how our own AMC standard (POL-AMC-001) answers each one. Use it as a yardstick for any contract, ours or anyone else's.
What is maintenance, and what is a change request?
The test we use is one sentence. If a request restores or preserves what the system already did, it is maintenance. If it changes what the system does, it is a change request.
That puts these inside the AMC:
- fixing defects in the scope that was delivered and accepted
- security patches and dependency upgrades
- server and uptime monitoring and backup checks, where we control the hosting
- configuration changes and minor tweaks under the effort threshold (below)
- restoring service after an incident
- advice and technical support
And these outside it, as separate project work: new features or modules, re-platforming and major version migrations, new third-party integrations, anything over the effort threshold, data entry, content writing, end-user training programmes, and hardware, licences or third-party subscription fees.
Patching belongs inside. NIST, the US standards body, describes patching as preventive maintenance and a cost of doing business. An AMC without dependency upgrades leaves a system that works today and gets harder to secure every month.
How fast should a problem be answered?
Severity should be set by business impact when the ticket arrives, not by how urgent the reporter's subject line sounds. Our standard uses four levels:
| Level | What it means | Response target | Resolution target |
|---|---|---|---|
| P1 Critical | Production down, a core function unusable for all or most users, confirmed data loss or an active breach, no workaround | 1 hour, 24x7, every day of the year | 8 business hours |
| P2 High | A major function degraded or unusable for some users, or a severe slowdown | 4 business hours | 3 business days |
| P3 Medium | A non-core function is faulty and a workaround exists | 1 business day | 10 business days |
| P4 Low | Cosmetic, documentation, or a question | 2 business days | Next maintenance window |
Read the clock rules under any table like this. Ours run in support hours only, except P1, which runs around the clock. The clock starts when the ticket lands in the agreed channel and pauses while we wait on the client for access, information or a decision.
Resolution targets are targets. Where the cause sits outside systems we control, such as a third-party API, a client-hosted server or an internet provider, the commitment becomes diagnosis, a workaround and escalation to that vendor. A workaround that restores service may come first, with the permanent fix in the next maintenance window. Nobody can promise a fix time for someone else's outage.
When is support open, and what happens out of hours?
Our support hours are Sunday to Friday, 09:00 to 18:00 Nepal Time (UTC+5:45), excluding Nepal public holidays. If you work a Monday to Friday week, note the Sunday start. The holiday list changes every year; the Ministry of Home Affairs publishes it on its government and public holidays page, and any AMC with a Nepal team should say which list applies.
P1 is answered at any hour. A P2, P3 or P4 raised out of hours counts as received at 09:00 the next business day. Out-of-hours work on anything below P1 needs written authorisation from your named representative first. Ask what that costs before you need it.
Intake matters as much as hours. We run one ticketing channel, and the ticket ID is the record. A P1 also needs a phone call to the on-call number. A WhatsApp message to one engineer, or a request made out loud in a meeting, is not a ticket and carries no service level. Each side names one primary contact and one escalation contact.
Where does a small tweak become a change request?
Under our AMC, any request estimated at more than 6 person-hours is a change request. It gets quoted in writing and approved before any work starts. Person-hours means total effort, not elapsed time: two developers for 3 hours is 6. Requests can't be split into smaller pieces to stay under the line. If the work is going to go over the approved estimate, it stops for re-approval. The full process is in our post on change requests in fixed-price projects.
The rule that protects you most is the reverse one. A defect in delivered, accepted scope is never a change request. It is fixed under the AMC however long it takes. When the two sides disagree about whether something is a defect or a change, the original written specification and the UAT sign-off decide. That is one more reason to run acceptance testing properly before you sign it off.
The AMC also includes a yearly bank of support hours. Time is logged in 15-minute steps against a ticket. The monthly report and review call don't draw from it, you are told in writing as it runs down, and P1 work never stops because the bank is empty. Unused hours expire at the end of the contract year; ask about expiry on any hour bank.
What does the client have to do?
An AMC is a two-way contract. Ours asks the client to:
- give timely access and credentials when a ticket needs them
- keep its own subscriptions, domains and payment methods current
- not change code, the database or server configuration without telling us
- provide a staging environment, or accept that fixes go straight to production
- test fixes within 5 business days
A contract with no client obligations makes its targets hard to enforce, because every miss can be blamed on something nobody agreed you would do.
Who owns the data and the code, and how do you leave?
Your data stays yours. We process it only to deliver the service. Access is through named individual accounts with least privilege and multi-factor login where the system supports it. When an engineer leaves the account, their access is revoked and secrets are rotated. A confirmed security incident affecting client data is reported to the client within 24 hours of confirmation. Nepal's Privacy Act, 2075 (2018) applies to personal information too; ask your own lawyer what it requires of you.
You always hold a current copy of the source code. It is never withheld, including during a payment dispute. Fixes made under the AMC belong to the client once paid for. Each month you get a service report: tickets by severity, results against the targets, hours used and left, incidents with their root causes, and patches applied.
On exit, there are up to 30 days of handover: source, credentials, documentation, your data in a standard format, and our access revoked with secrets rotated. That handover is never held back over a commercial dispute.
One more thing to check: when does the AMC start? Ours begins after a 90-day free defect-fix warranty from go-live. You shouldn't pay maintenance on bugs the build left behind.
What should you ask before signing any AMC?
- Is there a written list of covered systems: apps, environments, URLs, repositories and servers? Anything not listed is not covered.
- What is the one-sentence test for maintenance versus change?
- Who sets severity, and on what basis?
- When does the clock start, when does it pause, and does P1 run 24x7?
- What are the support days and hours, and which holiday list applies?
- What is the effort threshold for a change request, and can a defect ever be billed as one?
- What do you have to provide, and by when?
- Do you hold the source code now, and is it ever withheld?
- What happens to your data and access on exit, and how long is handover?
- Is there a warranty period before the AMC starts?
For the wider vendor checks, see how to choose a software development company in Nepal.
Frequently asked questions
Do I need an AMC if the software works fine today?
Probably. Dependencies and platforms keep moving, and security patches keep arriving. A system nobody patches doesn't stay fine; it just hasn't failed yet.
Can I get an AMC for software another company built?
Often, but expect an assessment first. Nobody should promise targets on code they haven't read.
What if the developer and I disagree on whether something is a bug?
Go back to the written specification and the UAT sign-off. If the system doesn't do what was specified and accepted, it is a defect. If it does, and you want it to do something else, it is a change.
Maintaining your system with us
Our business solutions team maintains the systems we build under this AMC standard, and the full terms are written into each contract along with the covered-systems list. If you have software that needs looking after, or an AMC you'd like a second opinion on, tell us what you're running.
Have a project in mind?
Let us help you build something scalable, fast, and built to last.
Start a conversationKeep reading
Travel agency software in Nepal: what Travel Quoter needed
Travel Quoter lets one traveller request reach several verified agencies. The four parts of that build, and what to plan for if you want software like it.
BusinessHow to choose a software development company in Nepal
How to choose a software development company in Nepal: check registration, open their live work, meet the builders, and get ownership and changes in writing.