Signal in the noise. From the first customer question to the merged fix.
Zygnal is an AI-native platform for support and engineering — living help docs, a learned support memory, tickets that link to your code, and an in-product drawer. One place where customer questions, support work, and engineering fixes all see each other.
Every resolved ticket becomes a captured insight. Insights become new articles and engineering fixes. Engineers triage tickets with code in context — soon, from inside their IDE. The loop closes.
Hosted in Sydney (ap-southeast-2). Multi-tenant from day one. Plugged into your code, not just your help desk.
Click Profile → Security, then Reset password. You'll get an email link that lasts 30 minutes.
Support tooling makes you choose. Zygnal does not.
Most teams duct-tape five tools together: a help-doc CMS, a ticket system, an AI chat add-on, a wiki for "things we have learned," and a project tracker the engineers actually open. Each one is partly aware of the others. None of them close the loop, and none of them know about your code.
Help docs go stale
You wrote them six months ago. The product changed twice. Now they're wrong, and customers find the wrong answers before they find your team.
Hard-won knowledge gets lost
Every resolved ticket teaches your team something. Most of that learning lives in someone's head, in Slack threads, or in a Jira comment nobody will read again.
AI without context is useless
Generic AI chatbots start cold every time. They do not know your product, your customers, or what your team has already worked out. So they hallucinate, or punt to human.
The fix loop never closes
Customer reports a problem. Engineer fixes it. Six weeks later, same question from a different customer. Nobody updated the docs. Nobody told marketing. The cycle repeats.
Four parts, one loop.
Your customers get a helpful AI right inside your product. Your support team gets tickets routed and triaged with full history. Your engineering team gets the ticket, the linked code, and the knowledge fed back as fixes and docs — soon, from inside their IDE.
Help docs that stay current
Markdown sources, AI-polished, continuously refreshed. Edit by hand or by source — Zygnal keeps a diff against the original. Each article carries page routes, freshness scores, and a "last validated" date so you always know what is fresh.
A learned support memory
Every resolved ticket becomes a structured insight: symptom, cause, resolution. Insights cluster into patterns. Patterns become new help articles and product-improvement candidates. The knowledge grows automatically.
Tickets, not a black box
First-class tickets with severity, priority, SLA, sprint backlog, and GitHub branch + PR linking. Threaded replies. Email-to-ticket via Mailgun. Sync with Jira. The canonical record lives with you, not in a vendor silo.
An in-product drawer
Embeddable widget your customers see inside your SaaS app. Contextual help, AI chat grounded in your knowledge, ticket creation, file uploads with vision OCR. HMAC-signed identity — no extra signup.
The closed loop:
Customer asks → drawer answers from KB → if not, opens ticket → AI triages with code from the linked repo → engineer ships a fix as a PR → merge updates the ticket, drafts the article, closes the loop with the customer → next customer finds it themselves.
The support platform that talks to your code.
Most "support tooling" treats the engineering hand-off as someone else's problem. Zygnal treats it as the point. Tickets link to repos. AI reads the code. Your IDE sees the prep. Fixes feed the knowledge base.
Or, more bluntly: prep is the bottleneck, not coding speed. Zygnal closes the prep gap.
-
Repository linking, scoped per product
Connect a GitHub repository to a product. Code access stays scoped to that product's repos — your support team can't accidentally read code they shouldn't.
-
Branch + PR linking on every ticket
One click creates a branch from your default base. PRs that reference the ticket update its status automatically via webhook. The link survives re-titling, re-assignment, and merging.
-
In-ticket AI with code in context
When triaging a ticket, the AI can read files, search code, and walk the directory tree of the linked repo — surfaced inline as the conversation happens. Every read is audited.
-
Sprints, severity, SLA
First-class sprint backlog with drag-to-rank, per-product calendars, and breach tracking. Engineering capacity sits next to support load, not in a separate Jira your team forgets to open.
-
Zygnal MCP server
A Model Context Protocol bridge that exposes ticket context, KB search, past resolutions, and code search to your IDE. Open Cursor, Claude Code, or Windsurf — your AI sees the full ticket prep without you copy-pasting context.
-
Code indexing for semantic search
Embeddings over your linked repos so the AI can find relevant code by intent, not just keyword. Becomes another retrieval surface alongside articles and resolution insights.
-
Closed-loop on PR merge
When the fix lands, Zygnal drafts the resolution insight, the article update, and the customer reply automatically. You review and ship.
Engineer's loop:
Customer reports a bug → support triages with AI, attaches the relevant code → engineer opens the ticket in their IDE via MCP → fix goes out as a PR → merge updates the ticket, drafts the article, closes the loop with the customer.
Everything support, nothing extra to wire up.
The features below are live in the platform today, in active internal use across our own product suite.
Hybrid retrieval
Full-text + semantic search, weighted and tunable. Markdown-aware chunking preserves heading context. Per-tenant + per-product partitioning — no cross-product bleed.
Insight capture & cluster mining
Every resolved ticket is summarised into a structured insight. Insights are clustered to surface emerging patterns and seed new article drafts.
Anomaly detection
Isolation-forest models watch ticket signals — volume spikes, unusual subject clusters, severity shifts. You see issues before they snowball.
Sprints & GitHub linking
Tickets sit in sprints. Engineers link a branch and PR with one click. Resolution closes the ticket and feeds the insight pipeline automatically.
GitHub App + in-ticket code access
Connect a repo per product. The triage AI can read files, search code, and walk the directory tree — scoped to that product's repos and audited end-to-end. MCP server for Cursor / Claude Code / Windsurf is coming.
Federation across products
Pull articles from existing help systems for unified search. Phase-1 customers federate with their existing knowledge base; nothing has to migrate on day one.
Email-to-ticket
Inbound email parsed and threaded as tickets via Mailgun. Replies from inside Zygnal go back out as email. The customer never sees a portal change.
Per-tenant SLA engine
Business calendars per tenant. Override profiles per product or per customer. Breach tracking built in. SLA is a first-class citizen, not a stitch-on report.
Audit log + multi-tenancy
Every mutating action is audit-logged via middleware. Tenant + product isolation enforced at the query layer, with row-level security in phase 2.
In-product drawer
HMAC-signed boot, JWT-scoped chat, contextual help by page route, ticket creation, file uploads with Vision OCR. Zero extra signup for your customers.
Per product, per month. No per-seat charge.
Drawer MAU per product is the value axis. Unlimited staff seats on every tier. 30-day free trial. Annual prepay saves 20%.
Base Zygnal — per product per month
Starter
Up to 2,500 drawer MAU
Core platform: KB articles, tickets, drawer with local AI chat, sprints, GitHub linking, federation, insight capture, full admin.
- Unlimited staff seats and end-customers
- Unlimited articles, tickets, insights, sprints
- Help portal with per-product branding
- Drawer with local AI chat (RAG-grounded)
- Email-to-ticket via Mailgun
- Jira import + webhook
- GitHub branch + PR linking
- Federation across products
- Email support
Growth
Up to 10,000 drawer MAU
Everything in Starter, plus the closed-loop intelligence: cluster mining, improvement drafts, anomaly detection.
- Everything in Starter
- Cluster mining + improvement drafts
- Anomaly detection on ticket signals
- Higher AI allowance
- Priority support
Scale
Up to 30,000 drawer MAU
For teams that need enterprise controls: SSO/SCIM, customer-managed keys, audit-log streaming, multi-region.
- Everything in Growth
- SSO / SCIM (Entra, Google, SAML)
- Per-tenant KMS keys
- Custom SLA
- Audit-log streaming
- Multi-region option
- Dedicated support
Zygnal Agent — add-on per product per month
Upgrade drawer chat from KB Q&A to a real support agent (Clara). Multi-step reasoning, search past resolutions, look up customer data, take actions across systems. Powered by KernelService under a white-label arrangement.
Starter Agent
Up to 500 chat-active MAU
Growth Agent
Up to 2,000 chat-active MAU
Scale Agent
Up to 6,000 chat-active MAU
Annual prepay
20% off list price.
Multi-product discount
2–3 products: −10%. 4+ products: −20%. Same tenant.
Free trial
30 days, single product, all features except SSO/KMS, 1k MAU cap. No card upfront.
Built for B2B SaaS, hosted where you'd expect.
Zygnal is being built for our own product suite first. The same posture you'd want for your customers' data is the posture we hold ourselves to.
Australian data residency
Hosted in AWS ap-southeast-2 (Sydney). Data does not leave the region by default. Multi-region available on Scale.
Multi-tenant from day one
Tenant + product isolation enforced at the query layer. Phase 2 adds Postgres row-level security and per-tenant KMS keys for Scale.
Audit log on every mutation
Middleware-driven audit logging on all writes. Stream to your SIEM on Scale. Nothing happens in the system that is not recorded.
HMAC-signed drawer identity
The in-product drawer authenticates customers via HMAC-signed boot tokens issued by your backend. Your drawer, your identity, no extra signup.
SOC 2 on the roadmap
SOC 2 Type II is a phase-2 commitment, scheduled alongside the commercial launch. Compliance posture is being built deliberately, not retrofitted.
Vendor-independent ticket data
Tickets and resolutions are first-class records in your tenant. Federation is one-way pull from external sources. You can leave; your knowledge comes with you.
Common questions.
Is Zygnal generally available?
What is a "drawer MAU" and what is a "chat-active MAU"?
How does the in-product drawer authenticate users?
How does Zygnal work with our existing Jira / GitHub / help system?
Is this just for support teams, or also for engineers?
How does the AI access code without leaking it?
Where is data hosted?
How is AI used? Is my data sent to third-party LLMs?
What is the difference between Zygnal and Zygnal Agent?
Can our engineers self-host?
How do you handle SLAs for your customers?
How do I get early access?
Be one of the first.
Zygnal is opening up to early-access customers as internal use stabilises. Tell us about your product and we'll be in touch when there's a slot.