Agent for Field Support

Grow your fleet without pulling engineers off the roadmap.

With Alloy, support resolves more tickets on their own, and engineers get to the root cause faster.

Read the film transcript

This robot is halfway through a job at a customer site. Then it freezes. The customer is left staring at a machine that will not move.

And the reason is buried across logs, support messages, and tribal knowledge. So an engineer gets pulled away from their roadmap to piece together that answer.

Alloy agent turns the whole incident into one investigation. Ask what happened, and it jumps to the failure, traces the signals around it, and shows the evidence behind the answer.

Support gets a clear next step. If engineering needs to step in, the relevant data is already attached. And the next incident starts with what the team learned from this one.

Help your engineers spend less time on support and more time on the core roadmap.

SwarmFarm · first-line support

~30% → 80%

of tickets resolved without engineering

SwarmFarm · time per ticket

1–2 h → ~10 min

for support to work a ticket

Andromeda · root cause

Weeks → hours

for engineering investigations

If you lead engineering

More robots in the field. More tickets on your engineers' desks.

A ticket comes in. The context support needs lives with one engineer, as tribal knowledge. So that engineer stops building to reconstruct what happened, while support and the customer wait.

  • Tickets pile up as the fleet grows
  • More engineers get pulled into support
  • The next generation keeps getting interrupted
With Alloy

Alloy helps support resolve more. And helps engineers get to the root cause.

Field ticketsFrom every robot in the fleet
alloy
SupportTicket resolved80%of tickets at SwarmFarm,
up from about 30%
EngineeringRoot cause fixedStarts from the evidence Alloy already gathered

Field tickets flow into Alloy. Most are resolved in support, about 80% at SwarmFarm, up from about 30%. The rest go to engineering with the evidence attached, to fix the root cause.

How it works

From ticket to next step, in the thread your team already uses.

Customers

Robotics teams already run support this way.

SwarmFarmalloy
"Last year at this time, I was spending a day or more investigating issues and support tickets. Now it's usually less than an hour."
Mitch GaleaSenior Autonomy Engineer, SwarmFarm

1 day → 1 hour

Mitch's time on support investigations: a day or more last year, usually under an hour now

A week

from connecting telemetry to first results

Read the SwarmFarm storyCase study · 9:00 film
SwarmFarm robots lined up on red soil, seen from above
CS-••••Robot stopped: lost comms with controllerInternal comment
Alloy Initial triage
  1. Summary. The robot lost communication with its main and attachment controllers mid-job and stopped. The time comes from telemetry.
  2. Priority and route. P2, robot stopped. Route to Robotics Support, per the triage guide.
  3. Related tickets. 2 earlier tickets with similar comms faults. Dropout at 4:23 pm.
  4. Logs. CAN log requested for 4:13–4:33 pm. ✓ Received
  5. Next steps. Review the CAN log for bus errors before contacting the farmer.
Telemetry windowonly this window pulled
3:00 pm3:304:004:30
Evidence

Don't trust it.
Check it.

Every finding links to the logs, tickets and code behind it. Correct it once, and the whole team gets the correction.

"It reinforces my team's knowledge and their confidence in their own calls."Jack Humphrey, Customer Success Team Lead, SwarmFarm
Knowledge base

Solve it once.

Alloy writes down every investigation, with the evidence behind it. Whoever picks up the next ticket starts from the answer. Your best AI setup remembers for one engineer. Alloy remembers for the whole team.

  • lf16_1014_1422.mcap
  • hil-thermal run #4812
  • ESC temps, fleet
  • KAN-15
  • #field-support thread

InvestigatingSolved once

ESC thermal margin on firmware 4.2

Above 34°C, ESC 3 limits throttle. Check it first.

ESC 3 hits 94°C at 00:14:32.

Fails on 4.2, passes on 4.1.

12 units above 90°C this week.

Customer report: LF16 lost thrust.

Maya saw this on LF11 in March.

GPS dropout on cold boot

Below 5°C, GPS needs 90 seconds to lock. Hold launch.

No fix for 91 s after power-on.

Timeout hard-coded at 60 s.

Still says 60 s is safe.

Three tickets closed as “retry”.

Dropouts cluster below 5°C.

Battery sag after overnight cold soak

Below 2°C, packs sag on first takeoff. Not a bad pack.

0.4 V sag on 31 cold starts.

Sag gone by second takeoff.

Pack health normal after warm-up.

Customer asked for 6 pack swaps.

Ruled out cell damage.

Comms loss past 2 km

Past 2 km, the old radio profile drops packets. Update it.

Link drops at 2.1 km, every flight.

Still on the 2025 power limit.

Packet loss jumps past 2 km.

Customer report: lost video at range.

Antenna ruled out in August.

MT
PN
NS
SO
LH
23times reused by
the team so far
Sample data
Integrations

Works with the tools and data you already have.

Not on the list? We'll build it with you.Forward-deployed engineers connect the rest of your stack, from custom robot APIs to in-house tools.Book a demo
Getting started

See value in week one. We do the integrations.

You grant access to your logs, tickets and code. We build and maintain the connectors, including custom formats. Answers show up in Slack and Jira.

Your team · Day one

Grant access

The only task on your side.

Alloy · Week one

Alloy investigates tickets

First pass on every ticket, with the evidence linked.

Alloy · Every week after

It gets faster

Known faults start from the last investigation.

Models, connectors and log formats are ours to maintain.SOC 2 Type 2 audited

Illustrative chart, no customer data

Still deciding?

Watch the whole story in 3 minutes.

Made forCTOs and Heads of Engineering at Series A–C robotics companies

Read the film transcript

If you lead engineering at a robotics company, you're trying to grow the fleet, improve reliability, and keep your engineers focused on building the product. But as more robots enter the field, more support tickets land on those same engineers' desks.

We work with many robotics teams, and we kept seeing the same problem. A ticket comes in, support searches for answers, but the context they need often lives with an engineer as tribal knowledge. So that engineer stops building the product to reconstruct what happened, while support and the customer wait.

As the fleet grows, the tickets pile up. More engineers get pulled into support, and work on the next generation keeps getting interrupted. So how do you grow the fleet without pulling more engineers off the roadmap?

Alloy gives support access to your robot data and what your team has learned from past issues, so they can resolve more tickets without escalating to engineering. And when a case does need engineering, Alloy gathers the evidence and presents its findings. Your engineer can review the case instead of starting the investigation from scratch. Less detective work, more time for engineering judgment, and for building the product.

At SwarmFarm, first-line support went from resolving approximately 30% of tickets to 80%, while time per ticket fell from one or two hours to approximately 10 minutes. At Andromeda, engineering root cause investigations went from weeks to hours.

Here's how that works. When a ticket lands in your support channel, Alloy searches the connected robot data, company context, and previous investigations. It brings its findings and supporting evidence back into the thread. Support can ask follow-up questions, understand the recommended next step, and work with the customer to resolve the issue.

With that context, support can resolve most cases without escalating to an engineer. When the case needs engineering judgment, the engineer joins the same investigation with the context already there.

Getting a robot working again and finding the underlying cause are different jobs. For a deeper investigation, the team can open the case in Alloy's platform to inspect logs, analyze telemetry, and run scripts.

As engineers and support teams investigate with Alloy, they add context, ask questions, and confirm what worked. Alloy retains what they learn, so each investigation builds on the last and the team's shared knowledge naturally compounds.

Ask Alloy a question in Slack. It's like asking your whole team across every time zone without interrupting anyone. Behind the answer is what your engineers and support teams have already learned together. Alloy draws on that shared knowledge, then brings the answer straight back into the conversation.

As more conversations and investigations add to that knowledge, the next person starts with more answers already available. They can get further without repeating work, then contribute what they discover for whoever comes next. The more your team uses Alloy, the more useful it becomes to everyone. With that history in one place, your team can identify recurring issues across the fleet.

Alloy connects to your existing tools and data through a suite of third-party integrations. For more custom needs, our forward-deployed engineers build integrations and workflows tailored to your team.

Resolve more in support. When engineering is needed, engineers can investigate more efficiently. Keep your team focused on building better robots. To see how Alloy could fit your team, talk to us.

What happens on the call?

Thirty minutes, and nothing to prepare. We learn how support runs on your fleet, show you Alloy on robots like yours, and agree what a pilot should prove.

  1. Your fleet today. Where field issues come in, and who gets pulled in to answer them.
  2. Alloy on robots like yours. A live investigation on data that matches your robots, not a slide deck.
  3. What a pilot would prove. Data access, security, and the one outcome you'd judge it on.
Where does our data go, and who can see it?

Your data stays in your own Alloy workspace, and you control our access to it.

We agree hosting with your team so it fits your security and compliance needs. We go through your security review with you: we share our SOC 2 Type II report, answer your security questionnaire, and list every provider we use in our Trust Centre.
What access do you need, and how much work is it for us?

Read-only access to where your data already lives. That's your team's part. Our forward-deployed engineers do the rest.

Alloy connects to your cloud storage, databases, Jira, Slack and Confluence. Nothing changes on your robots, and you don't change how they log.

~1 weekfrom connecting to first results at SwarmFarm

"When Alloy came on board, it was pretty much just, "Oh, give us access to your logs and these connectors." … it was generally pretty seamless."

Yan Chen, Co-founder and CTO, Andromeda

Why not build it ourselves, or point ChatGPT or Codex at our logs?

You can build a version for one engineer. Keeping it working for a whole support team is the hard part.

Claude Code or Codex wired to your logs works well for the person who set it up. Then support wants it too, and someone has to keep every connection alive, answer everyone's questions and hold on to what the last investigation found. Alloy is that layer, and it hands findings to your coding agent when it's time to fix the code.
"I could set something up and get it working well enough for me, but to support a whole team and having to then answer all their questions, that wasn't gonna be maintainable long term."

Mitch Galea, Senior Autonomy Engineer, SwarmFarm

Will it work with our robots and our stack?

Yes. It reads your logs as they are: MCAP, ROS, ROS 2, Parquet, CSV, ArduPilot and other formats.

You don't need ROS. If a format or tool isn't covered, our forward-deployed engineers build the connection for you.
Our robots are niche. Will it actually understand our domain?

It learns your robots from your own data, not from generic robotics knowledge.

Alloy works from your fleet's logs, tickets, docs and code. Our forward-deployed engineers tune it to how your team diagnoses issues, and your engineers' corrections add to its context in your workspace.
"There's some things that I've been doing for years as part of debugging, or you just know the answer to… It was just able to find it out in the first question."

Tom Brodie, Head of Autonomy, SwarmFarm

How do we know its answers are right?

Check them. Every finding links to the logs, tickets and code behind it.

Alloy posts its work for your team to review, and your people decide what goes back to the customer. You give it more room as trust builds.
"In the beginning, it was just kind of there in the background. We've stepping-stoned our way to it giving the answers forefront, so it's almost like it's been one of our engineers."

Sophie Lane, HRI and AI Engineer, Andromeda

What happens when it's wrong, or can't solve a ticket?

It escalates with the evidence, and it changes course when you correct it.

Engineering gets what Alloy checked, the signals around the fault and past tickets with the same signature. When a SwarmFarm engineer corrected a wrong assumption, Alloy withdrew the claim, reviewed related tickets for the same robot and proposed a better line of investigation.
Does it replace our support team?

No. It runs the first investigation on every ticket, so your team resolves more without escalating.

~30% → 80%of tickets resolved by SwarmFarm's L1 support

~200help desk conversations a week at Andromeda, investigated first by Alloy

What does it cost, during the pilot and after?

It depends on your fleet and how much you use Alloy, so we scope it with you after the first call.

Most teams start with a pilot on one workflow, then move to an annual partnership. We'll give you a business case to take to finance.
How do we know the pilot worked?

We agree the measures with you before the pilot starts.

For example: time to a first answer, tickets resolved before engineering, or engineering hours spent on support. At the end, you judge the pilot against them.

1–2 h → ~10 minper ticket at SwarmFarm

"Last year around this time, I was probably spending at least a day or more investigating issues and support tickets. But at the moment it's usually less than an hour here and there."

Mitch Galea, Senior Autonomy Engineer, SwarmFarm

Resolve more in support.Investigate faster in engineering.Keep your team building better robots.