For SaaS and developer-tools companies71% of tier-1 tickets resolved without a human
The AI support agent that unblocks your team for bugs and upgrades.
Aivastark is the AI support agent for SaaS products. Trained on your help center, API reference, and pricing. It resolves billing, onboarding, account, and integration tickets 24/7 · and routes only real bugs and edge cases to your engineers.
Replaces tier-1 support
Routes bugs to engineering via webhook
First reply: 14h to <2s
H
24/7Halcyon Support
AI support agent · <2s
H
Hi 👋 I'm Halcyon's support agent. I can help with billing, account, integrations, and the API.
How do I get a refund on my annual plan?
H
Yes. Annual plan▋
Ask about billing, API, account…
Powered by Aivastark
★ What your support agent handles
Three jobs your support inbox is choking on.
Most SaaS support volume is the same 50 questions: how do I, what's my plan, this endpoint is broken. Aivastark resolves the first two automatically and routes the third to the team that can actually fix it.
Docs deflection
Answers every "how do I…" question.
Most users won't read your docs. The support agent does, then answers them in seconds with a citation. Every reply links back to the source so they can dig deeper.
How do I rotate my API key?
→ API docs · §4.2
What's included in the Growth plan?
→ Pricing · features matrix
How do I enable SSO?
→ SSO guide · setup
Can I export conversation data?
→ Help center · exports
Cites every answer80%+ self-service
Billing & account
Resolves billing and account tickets instantly.
Charge questions, plan changes, seat upgrades, MFA resets, and trial extensions - answered against your fee schedule and account state, not invented.
Ticket received
"Why did my card get charged $60 instead of $48?"
Looking up the account
Growth plan · billed monthly · annual discount not active
Resolved with one reply
"$60 is monthly. Annual is $48/mo if billed yearly. Want to switch? Saves $144/yr."
Policy-aware · no fabricationMRR-saving conversations
Bug routing
Routes real bugs straight to engineering.
When the agent detects a bug it can't answer (stack-trace context, edge-case behavior, API 4xx), it captures repro details and pushes a structured payload to your webhook - ready for Linear, Jira, GitHub Issues, or Slack.
Webhook returning 502s since 2pm. Calls work in Postman.
Confidence 38% · capturing repro
Sorry about that. I'm gathering the details and routing to engineering now.Webhook fired200 OK
endpoint webhook 502 since 2pm
workspace halcyon-prod-eu
plan Enterprise
repro Postman OK · SDK fails
routed → Linear · ENG-3142
Linear · Jira · GitHub · Slack via webhook
Most SaaS teams have the support agent live by lunch.
No card required · React / Next.js SDK · Live in 5 minutes
Real support conversations
The tickets your support inbox sees every day.
Trained on your help center, API reference, pricing, and changelog. The agent resolves billing, onboarding, and developer questions in seconds, and escalates anything that needs a human with full context attached.
Billing11:42 PM
Why was I charged twice this month?
Looking at your account: one charge was your monthly Growth renewal, the other was a prorated seat upgrade you added on the 12th. I can email the line-item breakdown, or transfer you to billing if it still looks wrong.
Onboarding9:17 AM
How do I invite my team to the workspace?
Open Settings → Team → Invite member, then paste the emails. Growth includes 5 seats by default; let me know if you need more and I can show you how to scale the plan.
API / Developer1:03 AM
Where's the webhook docs? My events aren't firing.
Webhook setup is here: docs.halcyonworkspace.com/api/webhooks. Most fires-but-not-firing issues are HMAC signature mismatches · there's a 30-second debug guide on that page. If you've already checked, I'll capture the repro and route it to engineering.
Account access8:48 PM
Locked out of MFA, lost my recovery codes.
I can verify your identity via email and the last 4 of your billing card to reset MFA. This requires a human review · pulling in Lin from the account team now, expected reply within 15 minutes.
What you get
Built for the way SaaS teams actually run support.
Auto re-index
Stays in sync with ship velocity.
Every release, pricing change, and doc edit lands in the agent's brain within minutes. No annual "we should retrain the bot" project. Your support never goes stale.
2m agov3.4 release notes addedre-indexed
14m agoPricing page · Growth seats updatedre-indexed
1h agoAPI docs · webhooks v2 publishedre-indexed
3h agoSSO setup guide rewritere-indexed
Live · changelog-awarenever stale
Cost per ticket
Cuts cost per ticket from $2.50 to $0.60.
Median across SaaS teams running Aivastark at scale. Tier-1 deflection holds at 71%, with the saved capacity going back into onboarding, expansion, and engineering bandwidth.
Before
$2.50
per support contact
After
$0.60
per support contact
Tier-1 deflection71% · holding
Developer-friendly
Drops onto any modern stack.
React, Next.js, Vue, vanilla JS. Async loader is 18kb gzipped, CSP-safe, no iframes. Theme it with CSS variables. Identify the user via JWT so the agent sees plan, role, and account context.
One-line embed · any framework
<script src="cdn.aivastark.com/widget.js"
data-bot-id="acme-001"
data-user-jwt="{{user.jwt}}" async />React SDKNext.js componentVue pluginVanilla JSWebhook APIJWT identify
Enterprise
Passes procurement on day one.
SSO, audit logs, role-based access, SOC 2, GDPR, EU residency. The compliance answers your enterprise prospects ask before they sign · and the controls your security team needs after.
SSO · SAML/OIDC
Okta, Google, Azure AD.
Audit logs
Every agent action · S3 stream.
Role-based access
Admin · agent · viewer · custom.
SOC 2 Type II
Audit available on request.
GDPR DPA
On every plan.
EU residency
Frankfurt region · Enterprise.
Before & after
How SaaS support changes with an AI agent on duty.
Scenario
Without Aivastark
With Aivastark
"How do I rotate my API key?" at 2am
withoutStack Overflow answer · or a 14-hour wait
withCited reply from your API docs in <2s
Billing question · duplicate charge
withoutEmail thread, account lookup, manual reply 6 hours later
withAccount-aware reply, line items explained, resolved
Real bug · webhook returning 502
withoutGoes to inbox · CSM forwards to eng · eng asks for repro · 2 days
withRepro captured, structured payload to Linear in seconds
Trial user stuck on day 2
withoutChurns silently before anyone notices
withAgent answers the blocker · trial-to-paid saved
MFA reset request at 11pm
withoutEmail queue, escalation, next-business-day at best
withIdentity-verify flow started, agent on-call alerted
SSO setup question from enterprise prospect
withoutGoes to sales · sales asks SE · SE replies tomorrow
withCited reply from your SSO guide, deal velocity preserved
Spike during a launch
withoutHire seasonal CX, queue still slips, NPS drops
withSame flat bill, same <2s replies, no overtime
SaaS support teams running this in production
The numbers heads of support actually care about.
71%
tier-1 tickets resolved without a human
14h → 2s
first-reply on docs, billing, and API questions
76%
lower cost per support contact ($2.50 to $0.60)
80%+
self-service rate on the help center
"We deflected 71% of tier-1 tickets in the first week. My CX team got their nights back, my eng team stopped getting paged for docs questions."
Hanna Reid
Head of Support · Okra Finance
"The webhook routes bug reports straight into Linear with a clean repro. We stopped losing engineering hours to triaging support tickets."
Daniel K.
VP Engineering · Halcyon Workspace
"Onboarding completion went up because new users get unstuck instead of bouncing. Free-to-paid conversion is the part I didn't expect to move."
Elena Paredes
CX Lead · Mercator Pay
For SaaS support & engineering leaders
The questions SaaS support teams actually ask.
Will it actually deflect tier-1 tickets, or just deflect to a different inbox?
It resolves them. Aivastark replies in <2 seconds with a sourced answer pulled from your help center, API docs, or pricing page. Only the tickets it isn't confident about, or that match your escalation rules (refunds, MFA resets, bugs), reach a human. Median tier-1 deflection in production is 71%, with about 80%+ of help-center questions self-served end-to-end.
How does the agent stay current with our release velocity?
Aivastark continuously re-crawls your connected sources on a schedule you set · daily, hourly, or on-demand via webhook. When you ship a v3.4 release note or update pricing, the agent reflects it within minutes, not the next quarterly retraining batch. Most SaaS teams point it at their changelog feed so every shipped feature lands in support coverage instantly.
Can it actually identify the user and look up their account?
Yes. Pass a signed JWT via the `data-user-jwt` widget attribute and the agent sees plan, role, account ID, workspace, and any custom traits you encode. This lets it answer plan-specific questions ("is X included in my tier?"), unblock account-state issues ("why is my seat limit X?"), and route by tier (Enterprise issues fast-tracked to senior CX). Identity is verified, not just claimed.
How does it handle bug reports?
When the agent detects a real bug (4xx/5xx mentions, stack traces, edge-case behavior it can't resolve from docs), it captures the repro details · endpoint, request body, workspace ID, browser/SDK version, timestamps · and POSTs a structured payload to your webhook. Most teams pipe that into Linear, Jira, GitHub Issues, or a #bugs Slack channel. Engineers get clean tickets, not vague "help it's broken."
Does it work for developer-tools companies where users ask API/SDK questions?
Yes · this is a primary use case. The agent indexes your API reference and SDK docs alongside your help center. It handles endpoint questions, parameter explanations, auth setup, webhook configuration, rate limits, and SDK-version differences. If a user pastes a stack trace, it tries to match against known issues; if not, it escalates with the trace attached.
Can it work with Intercom, Zendesk, Help Scout, or Front?
Yes. Aivastark sits in front of (or alongside) your existing help desk. The widget runs on your app and docs site; escalations push to your help desk via webhook with full conversation context, so your existing team triages from the tool they already use. Migration is not required · teams typically run Aivastark as the tier-1 layer over their existing platform.
How is this different from Intercom Fin or Zendesk AI?
Intercom Fin and Zendesk AI are AI add-ons to those platforms, gated by their pricing model and locked into their stack. Aivastark is a standalone AI support agent that drops onto any product (React, Next.js, Vue, vanilla), trains on your real docs, routes wherever you want (your help desk, Linear, Slack), and prices flat per-month with no per-resolution surcharges. If you've outgrown your help desk's AI but don't want to migrate, Aivastark is the layer that fits over what you have.
Does it support multi-product or multi-tenant SaaS?
Yes. Each product gets its own agent with its own knowledge base, persona, and routing rules. Multi-tenant context flows through the JWT · the agent answers Tenant A questions from Tenant A's data, Tenant B from Tenant B's. White-label per tenant if needed (different brand name, colors, persona per workspace).
Will it answer in Spanish, German, Japanese, and other languages?
Yes. The agent runs on GPT-4o-mini (gpt-4o on Enterprise) and handles dozens of languages natively. It detects the user's language from their first message and replies in it. For international SaaS with global customer bases, this is one of the highest-impact features · non-English users no longer wait for business-hours in the right timezone.
Is it SSO-ready and enterprise-secure?
Yes. SAML and OIDC SSO via Okta, Google, Azure AD, or any IdP. Role-based access (admin, agent, viewer, custom). Audit logs streamed to your S3 or Datadog. SOC 2 Type II audit available on request. GDPR DPA on every plan, HIPAA BAA on Enterprise. EU data residency (Frankfurt) on Enterprise.
What about data security · does the agent see our customer data?
Only what you authorize through the JWT and what you put into the knowledge base. PII is redacted before any text reaches the LLM. Aivastark never trains shared foundation models on your conversations or docs. Embeddings are scoped to your organization only. On Enterprise, conversation data can auto-purge after 30 days.
How does pricing work compared to per-resolution AI tools?
Flat monthly: $20 Starter (1 agent, 1K conversations), $60 Growth (5 agents, 10K conversations, multi-channel inbox, auto-retraining), custom Enterprise (unlimited usage + named success engineer). No per-resolution fees, no per-seat creep. Volume spikes during launches or BFCM don't blow up the bill.
How long does setup take for a SaaS company?
Five minutes for a working agent on your docs site. A full production deployment with custom voice, JWT identity, webhook bug routing, and multi-channel coverage typically takes an afternoon. Free tier with no card required so you can pressure-test on real user questions before the engineering review.
Will it cannibalize our human support team's job?
It eliminates the repetitive tier-1 work ("how do I, what's my plan, where's the docs") and gives your humans more time for the complex, expansion-driving conversations they're actually good at. The teams running Aivastark report higher CSAT and better trial-to-paid conversion, not headcount cuts · the saved capacity goes into onboarding, expansion, and proactive outreach.
Most SaaS teams have the support agent live by lunch
Stop answering "how do I…" 9,000 times a month.
Your AI support agent is answering tier-1 in seconds and routing the real bugs to engineering · while your team ships.
React / Next.js SDK·SSO + audit logs·No card required