Open Source Helpdesk Software: Same Principles, Less Infrastructure

We believe in open source. We built Hay as open source. But open source and self-hosting are two different decisions, and conflating them costs teams more than it should.

The case for open source helpdesk software is strong and getting stronger. You get data sovereignty. You get transparency into how decisions are made. You're not locked into a vendor's roadmap. You never have to sit through another "we've updated our pricing" email from Zendesk. Developer communities on Reddit are full of teams who've built serious support operations on Zammad, Chatwoot, and osTicket. Their reasons are legitimate. We share most of them.

The deployment model is a separate question. Self-hosting means running your own infrastructure, which is straightforward if you've got the capacity in-house and meaningful work if you don't. For teams under 15 agents, the maths often favours managed. For larger teams with DevOps on staff, self-hosting often wins. Neither answer is about open source. Both are about team composition.

Why Open Source Matters for Support#

The frustration with proprietary helpdesk software is well-documented. Teams call it the "Zendesk tax": paying increasing amounts for basic ticketing features that feel like they should have been included from day one. Configuration options that require an upgrade. Reports that need an add-on. AI that's bundled at one tier and paywalled at another.

Open source sidesteps all of that. You can inspect the code. You can modify workflows that don't fit a vendor's template. You're not locked into a pricing structure that changes every renewal cycle. And you own the decision about where your data lives and which AI models process it.

The platforms people actually self-host reflect different priorities, and they're all credible:

Zammad is probably the most powerful of the three. Full-featured, modern interface, solid API. It needs Elasticsearch to run properly, which means your infrastructure requirements are higher than you'd expect. You're looking at 4GB+ RAM minimum for a production deployment, and that's before you add agents. But for teams with DevOps capacity, it's an excellent platform.

Chatwoot gets recommended a lot for teams that want mobile-friendly, multi-channel support. It's genuinely customisable and has a strong community. Native AI is more limited, so you'll likely bolt on third-party tools for intelligent routing and resolution, but the foundation is solid.

osTicket has been around forever. It handles basic ticketing well and has a massive install base. If your needs are straightforward (email ticketing, assignment, basic SLAs), it does the job with less overhead than the other two.

These are good tools. The question is how you deploy them.

Which Path Fits Your Team#

When teams evaluate self-hosted open source helpdesk software, they see the licence fee: $0. That's the part above the waterline. Everything below it (infrastructure, maintenance, engineering time) is what catches people off guard six months in.

Source: Qualimero

The maintenance line is the one worth sitting with. Two hours a month of engineering time at freelancer rates. Updates, patches, troubleshooting, the occasional "why is Elasticsearch consuming all the RAM" investigation. Reasonable work if you've got an engineer in-house. $2,640 a year if you're paying someone else to handle it.

Qualimero estimates the true monthly system cost of self-hosted open source at approximately €222, covering server costs, maintenance, and admin time. For a team of 3–5 agents, that often exceeds what a managed platform would cost. The crossover point where self-hosting becomes genuinely cost-effective is around 10–15 agents.

None of this is an argument against self-hosting. If you've got a DevOps engineer, a sysadmin, or a founder who genuinely enjoys running infrastructure, self-hosting Zammad or Chatwoot is a great call. The platforms are powerful, you own the stack end-to-end, and you can customise anything. For a 50-person support team with dedicated DevOps, it's often the right answer. For a team of 6 without that capacity in-house, managed tends to be the better fit. That's a team composition question, not an open source one.

If you've got an engineer who owns the stack, infrastructure time is productive work and the carrying cost is absorbed. If you don't, it shows up in moments like this: a Slack ping on a Thursday evening that Elasticsearch has eaten all the RAM on the helpdesk server, two hours of debugging instead of shipping the feature your customers have been asking about. Neither situation is wrong. They're different resource decisions, and the right one depends on who's on your team.

Every hour your engineering team spends maintaining helpdesk infrastructure, patching security vulnerabilities, and debugging deployment issues is an hour they're not spending on product or customer experience. That's fine if infrastructure is the work you want them doing. If it isn't, the path that fits is the one that takes it off their plate.

Where the Industry Is Actually Heading#

Zendesk's Matthias Goehler describes the shift well: AI in customer service is moving from simple automation to what he calls "anticipation." The goal has moved past answering tickets faster. It's about predicting what customers need before they ask.

The workforce side has been messier in practice. In 2023, Gartner's Drew Kraus predicted AI would reduce customer service agents by 20–30% by 2028. A Gartner survey of 321 CS leaders in October 2025 found only 20% actually reduced staffing. Gartner's Emily Potosky now predicts half the companies that cut staff due to AI will rehire by 2027. Support teams are being reshaped, not eliminated. The repetitive, low-judgment work shifts to AI. The complex, high-stakes conversations are becoming more important, not less.

That shift shows up in the metrics too. The old playbook measured speed: average handle time, first response time, ticket volume per agent. These are SLA metrics. They tell you how fast your machine runs.

The new playbook measures outcomes. Resolution confidence. Proactive resolution rate. Customer lifetime value. These are experience-level agreements (XLAs), and they tell you whether your customers are actually better off after interacting with support.

Sources: CMSWire, Auxis

This is where deployment model matters again. Most self-hosted open source helpdesk tools are built around the left column. They're ticket management systems (good ones), but fundamentally designed to move tickets from open to closed as efficiently as possible. Moving to the right column requires a different architecture: one where AI resolves conversations, where support data feeds back into product decisions, where the helpdesk is a strategic asset. Building and maintaining that architecture yourself, on top of managing the infrastructure, is a significant undertaking.

Open Source Without the Infrastructure Tax#

This is where Hay comes in, and we want to be clear about what it is: Hay is open source. The code is on GitHub. You can inspect it, fork it, contribute to it. What we've done is remove the deployment overhead so teams can get the transparency and flexibility of open source without running their own infrastructure.

The flexibility argument for self-hosting usually comes down to "I can modify the code to do what I need." With Hay, you configure AI behaviour through plain-English playbooks. Hit the "@" key and you're triggering real-world actions: refunds, order tracking, BI queries. You're writing instructions in language your whole team can read and modify. For the majority of teams, that covers the customisation they actually need, without forking source code.

The transparency argument is one we take seriously, because it's the reason Hay is open source in the first place. Autonomy Settings give you human-in-the-loop governance. You set the boundaries. The AI operates within them. Persona controls let you define tone and behaviour explicitly. You can see what the AI does and why, the same way you'd see it in the source code of a self-hosted tool.

Then there's vendor lock-in, the concern that draws a lot of people to open source. Hay is model-agnostic (Claude, GPT, DeepSeek), so you pick the model that fits your use case. If you've spent time in open source communities, you know how much the ability to choose matters. We built Hay around that principle.

And the practical side: Hay processes refunds through Stripe, tracks orders in WooCommerce, pulls customer data from HubSpot. 177 actions across 9 integrations. The AI resolves issues. It doesn't route them to a queue and hope someone picks them up.

Making the Decision#

If your team has the infrastructure capacity, self-hosting is a great call. Zammad, Chatwoot, and osTicket are genuinely powerful platforms, and running them yourself gives you full control over the stack: your data, your customisations, your upgrade schedule. For regulated industries, specific compliance requirements, or government contracts, it's often the only call that makes sense.

If you're on Chatwoot specifically and want an AI layer on top of your existing setup, there's a third path worth knowing about. One of our customers runs Chatwoot self-hosted for live chat and uses Hay as the AI layer, WhatsApp included. Their agents kept the tool they already trusted. Hay handles resolution underneath. Self-hosted where they wanted control, AI where they wanted velocity. We built the integration for them and it's running in production. Full documentation is still coming, so if that setup sounds like yours, get in touch and we'll walk you through it.

And if you don't have the infrastructure capacity in-house, managed open source covers the same principles without the pager rotation. You keep the transparency, the configurability, and the freedom to switch models. You skip the infrastructure line items.

For teams with the engineering capacity, self-hosted works well. For most ecommerce teams under 15 agents, managed is the cleaner fit. And for Chatwoot shops who want AI on top of what they already run, the third path is there.

Hay is open source and starts at €50/month. No per-agent fees on the Starter plan. Free 14-day trial, no card required. See if it fits →

References#

Auxis (2024) 8 IT Help Desk Metrics to Track for Optimal Performance. Available at: https://www.auxis.com/help-desk-metrics/ (Accessed: 13 March 2026).

Gartner (2023) Gartner Reveals Three Technologies That Will Transform Customer Service and Support By 2028. Press release, 30 August. Available at: https://www.gartner.com/en/newsroom/press-releases/2023-08-30-gartner-reveals-three-technologies-that-will-transform-customer-service-and-support-by-2028 (Accessed: 13 March 2026).

Gartner (2025) Gartner Predicts 50% of Organizations Will Abandon Plans to Reduce Customer Service Workforce Due to AI. Press release, 10 June. Available at: https://www.gartner.com/en/newsroom/press-releases/2025-06-10-gartner-predicts-50-percent-of-organizations-will-abandon-plans-to-reduce-customer-service-workforce-due-to-ai (Accessed: 13 March 2026).

Gartner (2026) Gartner Predicts Half of Companies That Cut Customer Service Staff Due to AI Will Rehire by 2027. Press release, 3 February. Available at: https://www.gartner.com/en/newsroom/press-releases/2026-02-03-gartner-predicts-half-of-companies-that-cut-customer-service-staff-due-to-ai-will-rehire-by-2027 (Accessed: 13 March 2026).

Goehler, M. (2025) Quoted in CX Today, "CX Trend Series 2025 Part One: Transforming Customer Service", 2 December. Available at: https://www.cxtoday.com/ai-automation-in-cx/2025-cx-trends-part-1-how-agentic-ai-is-set-to-deliver-on-decades-of-broken-promises/ (Accessed: 13 March 2026).

Goehler, M. (2026) "The Proactive CX Revolution: Using AI to Anticipate Problems, Not Just Solve Them", No Jitter, 2 February. Available at: https://www.nojitter.com/ai-automation/the-proactive-cx-revolution-using-ai-to-anticipate-problems-not-just-solve-them (Accessed: 13 March 2026).

Kihlström, G. (2021) "Create Better Customer and Employee Experiences With Experience Level Agreements", CMSWire, 18 October. Available at: https://www.cmswire.com/digital-experience/create-better-customer-and-employee-experiences-with-experience-level-agreements/ (Accessed: 13 March 2026).

Qualimero (2026) "Open Source Ticket System 2025: Complete Guide & Why Support Alone Isn't Enough". Available at: https://qualimero.com/en/blog/open-source-ticket-system-comparison-guide-2025 (Accessed: 13 March 2026).