Join the pilot
MenuClose

When your network is down or slow, know why, who’s affected and where.

Example: a fibre network with a cut cableExample diagram: light travels from an ISP’s OLT through fibre to homes. The cable between closures B4 and B7 is cut, the 14 homes behind splitter S-12 go offline, and SeruLabs marks the section, about 2.3 km from the OLT, and notifies a technician.OLTPON 3C2S-12S-11S-13S-14B4B7
Example: a fibre network with a cut cableExample diagram: light travels from an ISP’s OLT through fibre to homes. The cable between closures B4 and B7 is cut, the 14 homes behind splitter S-12 go offline, and SeruLabs marks the section, about 2.3 km from the OLT, and notifies a technician.OLTS-12B4B7

SeruLabs is an AI agent being built for Kenyan ISPs, on fibre and wireless. It’s designed to work out why customers are offline or slow, alert your team on WhatsApp, and find room for more customers.

Join the pilot See how it works

What SeruLabs is built to do

SeruLabs is built to work on fibre and wireless networks, from the first alarm to finding room for more customers.

Detect and diagnose

Know what’s wrong before anyone leaves the office.

  • Work out why customers are offline: a fibre cut, a power cut, a failed OLT port or radio, an upstream outage, a login problem or an unpaid account.
  • Tell a power cut from a cable cut, using the “dying gasp” signal fibre units send when they lose power, and Kenya Power’s published planned interruptions.
  • On wireless links, work out whether it’s a dish knocked out of line, interference on the channel, a blocked path, a sector with too many customers, or a radio that keeps rebooting because its power supply is failing.
  • List every affected customer and narrow the fault down to one splitter or section of cable, or one sector, link or tower.
  • When customers say the internet is slow, check whether it’s their connection, a full sector or PON, or your upstream link, and say when it’s more likely their own Wi-Fi.
  • Warn you before things fail: weakening signal, links that keep dropping and dishes drifting out of line.

Fix faster

Send the right technician to the right place, or fix it without a trip.

  • Send the on-call technician a trouble report on WhatsApp, with a map pin and the list of affected customers.
  • Suggest a fix your team can make remotely when one will do, like moving a sector to a cleaner channel.
  • Close a job only when the network shows service is back, with the technician’s photos and notes attached.
  • Check each new installation’s signal before the technician leaves, so rushed installs don’t turn into next month’s faults.

Keep customers

Tell customers what’s happening before they call you.

  • Tell affected customers on WhatsApp that you know about the problem, in English or Kiswahili, once your team approves the message.
  • Answer “Why is my internet down?” or “Why is it so slow?” on WhatsApp with the real cause, and tell customers with an expired package how to renew it by M-Pesa.
  • Warn customers the evening before a planned power interruption in their area.
  • Work out each customer’s real downtime and suggest a fair credit.

Capacity and revenue

Find room for more customers on the link you already pay for.

  • Show what fills your link at peak, by service, by sector and by heavy user.
  • Warn you when your link, a sector or a PON is close to full at peak.
  • Find the capacity you’re losing: expired accounts still online, speed profiles set above the plan, abusive traffic and unfair sharing.
  • Tell you how many more customers each sector or PON can take, and alert your waiting list when there’s room.
  • Estimate what local peering and caching would save you.
  • Catch billing mismatches: customers who’ve paid but are still blocked, and connections nobody is billing.

Before you buy more bandwidth

How it works

Connect your systems once. After that, SeruLabs is built to work through every outage or slowdown in the same order, the way a careful network engineer would.

  1. Detect. SeruLabs notices customer units going offline, links dropping, signals weakening or speeds falling.
  2. Group. Related alarms become one incident, whether it’s 14 units offline behind one splitter or 31 customers slow on one sector.
  3. Rule out power. It checks for the “dying gasp” signal fibre units send when they lose power, radios rebooting or going quiet together at one site, and planned Kenya Power interruptions in the area.
  4. Rule out billing. It checks that the affected accounts are active and, where plans are on record, on the right speed profile. It reads them from your billing system, RADIUS or the accounts on your routers.
  5. Locate. It compares affected customers by splitter, cable route, sector or tower, and reads signal and noise on the links, to narrow down where the fault is and what it is.
  6. Act. It sends the technician a trouble report on WhatsApp, or suggests a fix your team can make remotely, and drafts a notice for affected customers, for your approval.
  7. Close. It confirms from network data that service is back, then records the cause and how long the fix took.

Steps 1 to 4 are plain, predictable code. The AI handles the reasoning, the trouble report and the messages in steps 5 and 6. When the evidence is weak, it says so and recommends sending a technician.

Agent log

Example
  1. 19:42:07detectPON 3: 14 ONTs lost signal within 40 seconds
  2. 19:42:08groupGrouped into one incident, #1042
  3. 19:42:09powerNo dying gasp received. No planned Kenya Power outage listed for the area
  4. 19:42:09billing14 of 14 accounts active
  5. 19:42:10locateOLT port PON 3 up, transmit power normal
  6. 19:42:11locateAll 14 are behind splitter S-12. S-11, S-13 and S-14 normal
  7. 19:42:12locateRoute map: S-12 is fed through closures B4 and B7, about 2.3 km from the OLT
  8. 19:42:14locateProbable cause: fibre cut between B4 and B7 (confidence: high)
  9. 19:42:15actTrouble report sent to the on-call technician on WhatsApp, with a map pin
  10. 19:42:15actCustomer notice drafted in English and Kiswahili, waiting for your approval
  11. 1 h 36 min later
  12. 21:18:40closeLight restored on S-12. 14 of 14 ONTs back online
  13. 21:18:41closeIncident closed. Cause: cable cut during road works (technician photo attached)
  1. 19:56:03detectSector B, tower T2: 31 customers slow since 19:50, none offline
  2. 19:56:04groupGrouped into one incident, #1043
  3. 19:56:05powerSector radio up 41 days with no reboots. Not a power problem
  4. 19:56:05billing31 of 31 accounts active, speed profiles match their plans
  5. 19:56:06locateSector B airtime at 38%, so the sector isn’t full
  6. 19:56:07locateNoise floor on Sector B up from -92 to -78 dBm since 19:50. Sectors A and C normal
  7. 19:56:08locateSignal at the customer ends unchanged. Backhaul to T2 normal
  8. 19:56:09locateProbable cause: interference on Sector B’s channel (confidence: medium)
  9. 19:56:10actSuggested fix sent to your team: move Sector B to a cleaner channel. No technician trip needed
  10. 19:56:10actCustomer notice drafted in English and Kiswahili, waiting for your approval
  11. 9 min later
  12. 20:05:41closeChannel changed by your team. Noise floor back to -91 dBm, speeds normal on Sector B
  13. 20:05:42closeIncident closed. Cause: interference on Sector B’s old channel

What your customers see

Example message
Your ISP

Dear customer, we’re aware of an internet outage in your area, caused by a damaged fibre cable. Our technicians are already on the way, and we’ll message you as soon as service is back.

Mpendwa mteja, tunafahamu kuna tatizo la intaneti katika eneo lako, lililosababishwa na kebo ya fibre iliyoharibika. Mafundi wetu tayari wako njiani, na tutakutumia ujumbe mara tu huduma itakaporudi.

Dear customer, the internet in your area has been slow this evening because of radio interference. Our team is fixing it now, and we’ll message you as soon as speeds are back to normal.

Mpendwa mteja, intaneti katika eneo lako imekuwa na kasi ndogo jioni hii kutokana na mwingiliano wa mawimbi ya redio. Timu yetu inashughulikia tatizo hili sasa hivi, na tutakutumia ujumbe mara tu kasi itakaporudi katika hali ya kawaida.

Sent from your own WhatsApp Business number, once your team approves it.

Before you buy more bandwidth

At 9 pm the link is full and customers are calling to say the internet is slow, while people who want to join are told to wait. More bandwidth is expensive, so it’s worth knowing what’s filling the link first. Some of that traffic shouldn’t be there, and some customers get more speed than they pay for. SeruLabs is being built to find that capacity and put it to use, in five steps.

  1. See. In the first version

    Measure what fills your link at peak, by service, by sector or PON and by heavy user. Traffic is read only as totals by service, and a total per account, never what anyone browses.

  2. Recover. In the first version

    Find the free fixes. Look for expired accounts still online, connections nobody is billing, speed profiles set above the plan and queues that don’t share fairly at peak. Also look for heavy downloads that could move to a night package, and a backup link left idle at peak, where your contract allows using it. Recommend each fix; your team decides.

  3. Earn. In the first version

    Connect the customers who are waiting. Work out how many more each sector or PON can take, and message your waiting list on WhatsApp when there’s room, once you approve.

  4. Go local. In the first version, as an estimate

    Keep more traffic inside Kenya. Estimate how much of your peak could come from local peering and caching, and what that would save in shillings. Whether to set it up is your decision.

  5. Grow together. Later

    Share the cost of one connection to an internet exchange, or one cache, with other ISPs in your town, split by what each ISP uses, where the exchange’s rules and your licences allow it. It needs several ISPs on SeruLabs, so it comes later.

The free fixes make room for customers who are waiting. Their fees can help pay for local peering, which frees more room.

Capacity check

Example

Last 7 days

  • seeUpstream link full 21:00–22:45 on 6 of 7 nights
  • seeVideo 58% of peak traffic, software updates 11%, everything else 31%
  • seeThe 20 heaviest accounts use 24% of the peak
  • recover23 expired accounts still online at peak
  • recover9 connections online that nobody is billing
  • recover41 speed profiles set above the paid plan
  • earnRoom for about 30 more customers once these are fixed: 18 on Sector A, 12 on PON 2
  • earn11 people on your waiting list are in those areas. Message ready, waiting for your approval
  • go localAbout a third of peak traffic could come from local peering and caching

Ideas for later

  • Speed boosts customers can buy by M-Pesa, offered only when there’s spare room on their sector or PON.
  • Night and weekend packages tested against your real usage before you launch them.
  • Warnings a few days before big football nights and exam-results days, when traffic jumps.
  • An upgrade offer for customers whose slow internet keeps turning out to be their own router or Wi-Fi.
  • A reseller deal for customers who share their connection with neighbours, instead of only cutting them off.

Pilot ISPs get a free capacity check, as a one-page report on their own network.

Join the pilot

Other things it’s built to do

  • Trouble spots and history. Show where cuts and dropouts keep happening, for example along a route with road works, and how long repairs take.
  • Plain-language questions. Answer questions like “Which sectors were full at 9 pm last night?”
  • Your whole network. Work across GPON, point-to-point fibre and wireless links, on mixed equipment from different vendors.
  • Maps. Show the affected section, sector or tower on your map, and let technicians add to it from WhatsApp with photos.

For engineers

How SeruLabs is being built: where it runs, what it reads and where the AI fits in.

How SeruLabs connects to your networkDiagram: a collector inside your network reads whichever of these you run: OLTs, switches and routers, wireless radios and customer accounts. It has read-only access and only connects out, over an encrypted connection with a 4G backup, to the SeruLabs cloud. The cloud handles monitoring, diagnosis and capacity, alerts technicians and customers on WhatsApp or SMS, and keeps your team informed on the dashboard and WhatsApp.Your networkWhichever of these you runOLTsSwitches and routersWireless radiosCustomer accountsCollectorRead-only4G backupOutbound only,encryptedSeruLabs cloudMonitoringDiagnosisCapacityDashboardTechnicians andcustomers(WhatsApp, SMS)Your team(dashboard andWhatsApp) How SeruLabs connects to your networkDiagram: a collector inside your network reads whichever of these you run: OLTs, switches and routers, wireless radios and customer accounts. It has read-only access and only connects out, over an encrypted connection with a 4G backup, to the SeruLabs cloud. The cloud handles monitoring, diagnosis and capacity, alerts technicians and customers on WhatsApp or SMS, and keeps your team informed on the dashboard and WhatsApp.Your networkWhichever of these you runOLTsSwitches and routersWireless radiosCustomer accountsCollectorRead-only4G backupOutbound only,encryptedSeruLabs cloudMonitoringDiagnosisCapacityDashboardTechnicians andcustomers(WhatsApp, SMS)Your team(dashboard andWhatsApp)

A small collector runs inside your network.

On a mini-PC, a virtual machine or a spare server in your equipment room. It reads your equipment and customer accounts with read-only access.

Nothing gets in.

The collector only ever connects out to SeruLabs. You don’t open ports or set up a VPN, and nobody can reach into your network through it.

Alerts still go out when your internet doesn’t.

A backup 4G SIM keeps alerts flowing when your own link is down. If the connection drops, the collector stores readings and sends them once it’s back.

Where the AI fits in.

Polling equipment, grouping alarms, spotting patterns like dying gasps and adding up traffic is ordinary, predictable code. The AI, Claude by Anthropic, runs when there’s something to explain. In an incident, it weighs the evidence, writes the trouble report and drafts customer messages. For capacity, it writes the report from the week’s numbers.

Your team stays in control.

Customer messages need your approval. SeruLabs doesn’t cut anyone off or change your network itself: it recommends, and your team decides and makes the change. When the evidence is weak, it says so and recommends sending a technician.

Official WhatsApp, your own number.

Messages go through Meta’s official WhatsApp Business Platform, with SMS as a fallback. Customers hear from your WhatsApp Business number, not ours.

What it reads

GPON fibre
ONT status, optical power levels, dying-gasp alarms, OLT port alarms and traffic on each PON port
Point-to-point fibre
Port status, optical readings and error counters on switches and routers
Wireless links
Signal at both ends, signal-to-noise ratio, noise floor, sector load, link capacity and reboots, through tools like UISP, the MikroTik API or Cambium’s management platform
Customer accounts
Sessions, speed profiles and account status from RADIUS, your billing system or the accounts on your routers. Plans and payments, including M-Pesa, from your billing system
Traffic
Totals by service, and a total per account, from your routers’ flow data, never individual browsing

Security and data

You’re trusting SeruLabs with your network and your customers’ details. Here’s how it’s being built to handle them.

What it collects
Network alarms, signal readings, equipment status, traffic totals by service and a total per account, session records, customer account IDs, plans and payment status, and the phone numbers of customers and people on your waiting list, used only to send them messages
What it never collects
What your customers browse, or the content of their traffic
How it’s protected
Read-only access, a collector that only connects out, encrypted connections, and each ISP’s data kept separate from every other ISP’s
Who owns it
You do. Your data is deleted when you leave
Agreements
Before a pilot starts, we sign a data processing agreement with you

Coming next: automatic fault location

In development

The first version of SeruLabs is built to tell you a cut has happened and which section of your network it’s in. Next, we’re adding a low-cost testing unit that works with the agent. When a cut is detected, it automatically runs an OTDR test, the same technology fibre technicians already trust, and measures how far along the line the break is. Matched against your route maps, that sends your team straight to the right section, without a test trip first.

  • Built on proven OTDR fibre testing
  • Runs automatically the moment a cut is detected
  • Results appear in your SeruLabs dashboard and WhatsApp alerts
  • Starting with feeder, backbone and point-to-point fibre, where no splitter is in the way
  • For fibre only: wireless links are diagnosed from their radio readings

Pilot partners get first access.

Join the pilot
Example trace Example OTDR traceExample OTDR trace on a point-to-point fibre: the trace slopes down gently along the fibre, with small steps at splices, then drops sharply at a break 2.62 km from the start.0123Distance (km)Loss (dB)Break: 2.62 km Example OTDR traceExample OTDR trace on a point-to-point fibre: the trace slopes down gently along the fibre, with small steps at splices, then drops sharply at a break 2.62 km from the start.0123Distance (km)Loss (dB)Break: 2.62 km

Join the pilot

We’re looking for our first partner ISPs in Kenya, on fibre, wireless or both. The pilot is free, and you’ll work directly with Seru, the founder.

What you get

  • A free capacity check to start: we collect a week of readings from your network, read-only, and send you a one-page report on what fills your peak, what’s being wasted and how many more customers you could connect.
  • SeruLabs on your network for the length of the pilot, at no cost.
  • A real say in what gets built next.
  • If you run fibre, first access to automatic fault location.

What we ask

  • Read-only access to your systems.
  • Honest feedback, good and bad.
  • If SeruLabs works for you, a testimonial in your own words.

Before we start

We agree the pilot’s scope in writing and sign a data processing agreement with you.

Tell us a little about your network.

Email your details to hello@serulabs.com.

Network type (optional)

We need a phone number or an email address to reply.

This opens your email app with your details filled in, addressed to hello@serulabs.com.

Prefer WhatsApp? Chat with Seru on WhatsApp

Built by a former ISP technician

Emmanuel Seru Shikoli, founder

“A customer would call to say their internet was down, and we’d spend hours just finding where the problem was. SeruLabs is the tool I wish we’d had.”

I spent two years as a technician at an ISP, doing the day-to-day work of keeping customers connected. I also build AI agents, and SeruLabs brings the two together: an agent that does the searching, so ISP teams can spend their time fixing.

We’re starting with ISPs. The same approach, an agent that connects to a business’s systems, finds what’s wrong and handles routine work, will extend to other industries like schools and hotels.

Built in Kenya.

Questions

How is SeruLabs different from the monitoring tools I already use?

Tools like The Dude, Zabbix, LibreNMS and UISP tell you that something is down. SeruLabs is built to tell you why, who’s affected and where, then act on it: a trouble report for your technician and a notice for your customers. It’s also built to show what’s filling your link at peak and how many more customers you could connect, and to work alongside the tools you already run.

My network is mostly wireless. Is SeruLabs for me?

Yes. SeruLabs is being built for wireless links as much as for fibre. It’s designed to read signal, noise and load on your sectors and backhaul links, and to tell a dish knocked out of line from interference or a sector that’s simply full. Automatic fault location is the one part that’s for fibre only.

What if I don’t use RADIUS or a billing system?

That’s fine. SeruLabs is built to read customer accounts from RADIUS, a billing system or the accounts on your routers. Checks that need to know what each customer pays for, like speed profiles set above the plan, need a record of plans, so at the start of a pilot we agree where that comes from.

Do you need access to my network?

Read-only access, through a small collector that runs inside your network. It only ever connects out to SeruLabs, so you don’t open ports or set up a VPN, and nobody can reach into your network through it.

Does SeruLabs change my router settings?

Not on its own. It recommends changes, like cutting off an expired account or correcting a speed profile, and your team approves and applies them.

Do you see what my customers browse?

No. Traffic is analysed only as totals by service, like video or software updates, and a total per account, never individual browsing.

What does the free capacity check involve?

Your team turns on flow export on your router if it’s off, and we help install the collector. Then we collect a week of readings from your routers, radios, OLTs and customer accounts, read-only, and send you a one-page report on what fills your peak, what’s being wasted and how many more customers you could connect. Apart from that, nothing on your network changes, and traffic is read only as totals by service, and a total per account. It starts once we’ve signed a data processing agreement with you.

Where is my data stored?

In the SeruLabs cloud, kept separate from every other ISP’s data. Before a pilot starts, our data processing agreement sets out exactly where it’s hosted, how long it’s kept and how it’s protected, including any transfer outside Kenya.

Which equipment does it work with?

It’s built for mixed networks: GPON OLTs, point-to-point fibre switches, routers such as MikroTik, RADIUS or billing systems, and wireless links managed through tools like UISP, the MikroTik API or Cambium’s platform. Most mainstream OLTs and radios report what SeruLabs needs, but cheaper brands vary, so we check your equipment at the start of a pilot.

What happens when the cause isn’t clear?

SeruLabs says so. Instead of guessing, it tells your team what it checked and what it found, and recommends sending a technician.

Will automatic fault location work beyond splitters?

Not at first. We’re starting with feeder, backbone and point-to-point fibre, where no splitter is in the way and testing from your equipment room is most reliable, and expanding from there.

How does pricing work?

A simple monthly subscription based on your number of active customer lines, paid in shillings by M-Pesa or bank transfer. Pilot partners start free.

Can I use SeruLabs today?

Not yet. SeruLabs is in development, and we’re looking for our first pilot ISPs now. Join the pilot to be one of them.