Honest, per-product pricing.

Not yet. Zygnal is in beta, currently used internally across the team’s own product suite. We are taking early-access requests now and will open up self-serve signup in phase 2 once internal use validates the value.

Drawer MAU is a unique end-customer who opens the in-product drawer at least once in a calendar month, scoped to your tenant + product. Chat-active MAU is the subset who actually starts a chat session — opening and bouncing does not count. Both are visible live in the dashboard.

Your backend issues an HMAC-signed boot token containing the customer identity. The drawer presents that token; Zygnal verifies the signature and scopes the session. There is no separate signup, no shared secrets in the browser, and no Zygnal account required for end-customers.

Tickets sync with Jira via webhook + import. GitHub branches and PRs link directly to tickets. Existing help articles can be federated in (one-way pull) for unified search before you migrate any content. Nothing has to move on day one.

Both. Today, engineers get one-click branch creation from tickets, PR-state webhook sync, and an in-ticket AI that can read files and search code from the linked repo while triaging. Coming next: a Zygnal MCP server so Claude Code, Cursor, and Windsurf can pull the full ticket context — KB articles, past resolutions, and code search — directly into the editor. Prep is the bottleneck, not coding speed.

Code access goes through our GitHub App, scoped per product to the repos your team explicitly links. The triage AI uses three tools: read a file at a ref, search code, and list a directory — every call is recorded in an audit log with who, what, and why. No code is stored in the AI provider or used for training.

AWS ap-southeast-2 (Sydney). Data does not leave the region by default. Scale tier customers can opt into multi-region deployment. Per-tenant customer-managed KMS keys are available on Scale.

AI is used for chunk summarisation, embedding generation, and drawer chat. We use Anthropic and OpenAI under enterprise zero-retention agreements. Your content is not used to train external models. The agent layer (Zygnal Agent add-on) is powered internally by KernelService — invisible to your customers.

Zygnal (base) gives you a drawer with grounded KB chat — answers from your help articles, with sources. Zygnal Agent upgrades that drawer to a multi-step agent (Clara): it can search past resolutions, look up customer data, and take actions across systems. You can subscribe to base only, then add the agent later.

Not in phase 1. Enterprise self-hosting may be considered case-by-case in phase 2 as part of a Scale-tier conversation, but the product is built and operated as a hosted SaaS.

Per-tier SLAs published with the contract. Internally, Zygnal has a per-tenant SLA engine — business calendars, override profiles per product or customer, breach tracking. We use it on ourselves first; what we promise externally is what we operate against internally.

Request early access from any "Get early access" button on the site. We are gradually opening Zygnal to external customers as internal use stabilises. Early-access customers get priority onboarding and locked-in beta pricing.