Support Agent
Support Agent is an AI support automation platform from Estonian studio AI Lab. It builds visual workflows where AI agents reason through requests, call tools to act on business systems, and escalate to human staff only when genuinely needed.
What is Support Agent?
Support Agent is a customer support automation platform published by AI Lab, the product brand of Estonian company Kempu OÜ. AI Lab is primarily an AI integration consultancy, and Support Agent is one of its three products alongside Publisher, a content platform, and Play Text, a free text-to-speech app.
The product is pitched against two failure modes the vendor names explicitly: chatbots that deflect volume but cannot take action, and rule-based automation that triggers reliably but cannot reason through ambiguity. Support Agent's answer is to combine both in one deployable unit.
Workflows are built in a Visual Flow Designer, where each process is a directed graph of typed steps — Filter, Decision, Validation, Transform, Action and Agent — joined by conditional edges with prioritised routing. A flow may run up to 100 steps and ships with loop protection, and each invocation chooses synchronous or asynchronous execution.
Within a flow, each agent is an independent reasoning unit. It receives inputs, works through the problem, calls tools in a full agentic loop and returns a result, without the inner loop being orchestrated by hand. System and user prompts accept injected variables, and model, temperature, token limit and timeout are set per agent. Agents can switch between Anthropic, OpenAI and Google individually, which the vendor presents as avoiding provider lock-in without rewiring workflows.
Connection to existing systems runs through a plugin ecosystem. A single plugin can expose tools for agents, actions for flows and triggers that start flows; plugins are auto-discovered, can run as multiple instances under different credentials, and carry health checks.
Operations are covered by observability: logging at run, step and tool-call granularity, live metrics for active agents, running flows and success rates, and transparent cost tracking down to the token. Human escalation is built in and carries full context, workflow logic is kept separate from response templates so wording can change without touching flows, and language-aware retrieval with automatic fallback serves several languages from one workflow.
It deploys either on AI Lab's managed cloud or self-hosted on the customer's own infrastructure. Published use cases span hosting and SaaS, e-commerce and enterprise IT. The product is sold through a demo request: there is no public pricing, trial or sign-up.
What it does
- Run AI agents that reason through ambiguous requests and decide what to do next
- Call tools that read from and write to your existing business systems
- Orchestrate multi-step workflows as a typed, conditionally routed flow graph
- Escalate to a human operator with full context when the agent reaches its limit
- Log every decision, step and tool call for a complete audit trail
- Meter token usage and cost per run, per step and per tool call
- Serve several languages from a single workflow with automatic fallback
When to use Support Agent / When not to
A quick filter to help you decide if Support Agent is the right fit.
When to use Support Agent
- Support teams that need automation to take real action on internal systems, not merely deflect tickets with a chatbot
- Hosting and SaaS providers automating billing lookups, end-to-end password resets and ticket triage
- E-commerce operations handling order status, shipping updates and refund eligibility against business rules
- Enterprise IT and service desks routing VPN, software and HR requests, and syncing incidents with an ITSM tool
- Organisations with data residency or auditability constraints, who need self-hosting and a full decision trail
When not to use Support Agent
- Buyers who need a published price, since no rate card exists anywhere on the site and every path leads to a demo request
- Teams wanting to self-serve, as there is no free plan, no trial and no online sign-up
- Anyone requiring contractual guarantees up front: no terms of service, privacy policy or DPA covers this product
- Organisations whose procurement demands published security certifications, as none is claimed
- Developers hoping to integrate through a documented public API, which is not published
How to use Support Agent
A typical end-to-end flow, from setup to results.
- Request a demo through the contact form, as there is no self-service sign-up
- Decide on a deployment mode: AI Lab's managed cloud, or self-hosted on your own infrastructure
- Map the support process you want to automate as a flow of typed steps in the Visual Flow Designer
- Add conditional edges with prioritised routing so the flow branches on your business rules
- Configure each agent: system and user prompts, injected variables, model, temperature, token limit and timeout
- Choose the LLM provider per agent among Anthropic, OpenAI and Google, using your own provider accounts
- Install the plugins that connect to your existing systems, then sync so their tools and triggers are discovered
- Set credentials per plugin instance and confirm connectivity with the built-in health checks
- Define the escalation rules that hand a case to a human operator with its full context
- Monitor runs, success rates, step-level logs and per-token cost in the observability console
Pros & Cons
Pros
- Agents act on real systems, reading and writing through tools rather than only generating replies
- Self-hosting gives full control over where data physically lives, with no SaaS intermediary
- No LLM vendor lock-in: three providers are interchangeable at the level of the individual agent
- Fine-grained traceability, with every decision and tool call logged and inspectable
- Cost is visible per token and per run, which makes AI spend budgetable rather than a surprise
- Flows are designed visually, so business logic can be expressed without writing code
- The publisher is an identifiable EU company, verifiable in the Estonian register and in VIES
Cons
- No pricing is published anywhere, so the budget cannot be assessed without talking to sales
- No free plan, no trial and no self-service sign-up: the only entry point is a demo request
- No terms of service, privacy policy or DPA covers this product
- No public API or plugin documentation, despite the platform being extensible by design
- No security certification is claimed, and no subprocessor list is published
- Multilingual support is promised but not one language is actually named
- No named customer, usage metric or case study substantiates the product's claims
Pricing & Plans
No price is published for Support Agent. A complete enumeration of the site confirms that no pricing page exists, and every commercial path leads to a demo request rather than to a rate card. There is accordingly no free plan and no free trial on offer for this product, and the starting price, currency and billing unit are all left empty rather than estimated. Prospective customers should also note that the large language model consumption sits outside any vendor fee: the platform runs on the customer's own Anthropic, OpenAI or Google accounts, and meters that spend per token and per run. AI Lab separately advertises a complimentary 30-minute technical consultation, which belongs to its consultancy activity rather than to this product.
- Cloud — deployment managed by AI Lab
- price not published
- Self-Hosted — deployment on the customer's own infrastructure
- price not published
Data, GDPR & hosting
A consolidated view of how Support Agent handles your data.
GDPR overview
The publisher, Kempu OÜ, is established in Tallinn, Estonia, and holds intra-community VAT number EE102625590, valid in VIES as of 1 September 2026, so the GDPR applies to it by law. Beyond that, the site makes no GDPR statement at all for this product: it neither claims compliance nor denies it. There is no privacy policy covering Support Agent, no data processing agreement published or offered on request, no named data protection officer and no dedicated privacy contact. An Article 27 representative is not required, the publisher being established inside the Union. The contact form is protected by Google reCAPTCHA, whose own privacy policy and terms are referenced. The self-hosted deployment mode is the only concrete lever offered over where personal data physically resides.
Who owns the data?
No published contract governs Support Agent. A complete enumeration of the site returns ten pages, and the only legal document among them covers Play Text, a separate consumer app from the same publisher, not this platform. Data ownership is therefore never stated in writing. The one substantive signal is architectural: in self-hosted mode the vendor positions the deployment as running on the customer's own infrastructure, with no SaaS intermediary and full control over where data lives, which in practice leaves the customer holding the data. For managed cloud deployments nothing equivalent is documented. Anyone contracting for this product should expect ownership, licensing and processing terms to be settled in a negotiated agreement rather than read off the website.
Reuse rights
The site sets out no reuse rights, no licence grant and no restriction on what either party may do with the data flowing through the platform, because no terms of service exist for it. What can be observed is operational rather than contractual: the platform logs every run, every step and every tool call, and meters token consumption and per-run cost, so execution data is certainly recorded. Model processing depends on the LLM provider the customer selects among Anthropic, OpenAI and Google, and is governed by that provider's own terms and the customer's own account with them. Whether the vendor may reuse customer data is never addressed in either direction, so it should be negotiated explicitly.
Data retention & training
Hosting summary
Support Agent is offered in two deployment modes, and they differ sharply on hosting. In self-hosted mode the platform runs on the customer's own infrastructure, which the vendor frames as your infrastructure and your data, with no SaaS intermediary and full control over where data lives; jurisdiction is then whatever the customer's own hosting arrangement dictates. For the managed cloud mode, no hosting country, region or provider is named anywhere on the site, and no data residency commitment is published. Note that the ailab.ee marketing site itself resolves to Amazon infrastructure in Dublin, Ireland, but that is the WordPress site, not the platform, and it should not be read as the product's hosting location. A further consideration is that model inference is carried out by whichever third-party provider the customer selects — Anthropic, OpenAI or Google — under the customer's own account, so part of the processing footprint sits with that provider rather than with AI Lab.
Things to keep in mind
Risks and trade-offs to weigh before adopting Support Agent.
- The site's only privacy policy covers Play Text, a sibling app, not this platform: do not read it as governing Support Agent
- No terms of service or DPA exists for this product, so data ownership, reuse and retention are entirely a matter for negotiation
- Agents write to live business systems, so a mis-specified flow can act wrongly at scale rather than merely answer wrongly
- The publisher was incorporated in January 2019, but the AI Lab brand is a 2025 development and the entity previously traded under other names
- The publisher is a very small company, which is a continuity consideration for a platform embedded in support operations
- LLM spend sits on your own provider accounts and scales with volume, so the true running cost is not the licence alone
- Over-automating support can erode the tacit product knowledge a human team accumulates from handling awkward cases itself
Setup & Integrations
Technical difficulty
Moderate to high, and not self-service. Flows are designed visually without code, but onboarding begins with a vendor demo rather than a sign-up. Self-hosting is described as a single deployable unit, which still assumes an operations team. Connecting existing systems means installing plugins and configuring credentials per instance, and agents need their prompts, model, temperature and limits tuned. Customers must also bring their own Anthropic, OpenAI or Google accounts.
Deployment
Integrations
Behind Support Agent
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
How is Support Agent different from a chatbot?
Can I host it on my own infrastructure?
Which AI models does it use?
How much does it cost?
Is there a free trial or a way to sign up online?
What happens when the AI cannot resolve a case?
How auditable is it?
How large can a workflow be?
Does it offer a public API?
Who publishes Support Agent?
Should you pick Support Agent?
Support Agent is a coherent and unusually specific piece of engineering writing. Its product page describes a real architecture rather than a slogan: typed flow graphs with conditional routing, agents whose model and parameters are set individually, a plugin model that cleanly separates tools, actions and triggers, and logging at three levels of granularity with a cost attached to every run. The two design choices that stand out commercially are self-hosting, which puts data residency entirely in the customer's hands, and per-agent switching between Anthropic, OpenAI and Google, which is a genuine hedge against provider lock-in.
The reservations are about everything surrounding the product rather than the product itself. There is no published price, no trial and no way to sign up, so evaluation cannot begin without a sales conversation. More significantly, no contractual document covers this platform at all: the site's only legal page belongs to a sibling consumer app, and there is no terms of service, privacy policy, DPA, subprocessor list or security certification to read. The multilingual claim names no language, and no customer, metric or case study is offered as evidence.
The publisher is verifiable, which counts for something: Kempu OÜ is a real Estonian company, confirmed in the national register and in VIES, though a small one, reporting €33,366 of revenue for 2025. It is also worth knowing that the AI Lab brand is recent even though the legal entity dates from 2019.
Support Agent suits an organisation prepared to run a scoped, vendor-supported project and to negotiate its own contractual terms. Anyone expecting to compare prices, self-serve a trial or rely on published compliance documentation should look elsewhere.
- Choosing a selection results in a full page refresh.
- Opens in a new window.