Krina — Mentions and Follow-ups for Jira and Confluence

Sivect · Krina

Krina by Sivect is a Forge-native app that captures every @mention across Jira, Jira Service Management, and Confluence, classifies each one with AI across three dimensions, and surfaces them in a prioritised personal inbox so nothing falls through the cracks.

Researchers identified this exact failure mode in 2003 — work lost because the communication channel and the task-management layer were separate systems. Krina unifies them.

— Bellotti et al. · ACM CHI 2003

Get it on Atlassian Marketplace
Free to try · for Jira · Confluence · JSM
Forge-native
GDPR Art. 25
Atlassian Cloud
Data residency preserved
Deadline-aware urgency
Krina
Live
All 5Jira 3JSM 1Confluence 1
ARCH-1v4.2.1 hotfix release
ApproveHigh · NowCustomer Due in 4 days

@you release checks all green in staging. Waiting for your approval to proceed with prod deployment.

SCSarah Chen · 2h agoResolve
PEOPLE-4Headcount approval — Q3
RespondHigh · TodayTeam Budget window closes tomorrow

@you need your formal headcount request by COB — Finance closes Q3 budget tomorrow morning.

3D AI Classification · Action · Urgency · Impact87ms

RUNS ON ATLASSIAN — WHAT THIS MEANS FOR YOU

  • Your data never leaves Atlassian. All compute and storage runs inside Atlassian's own cloud — the same environment your Jira and Confluence data already lives in. No external vendor servers. No new security review.
  • Data residency included. If your Atlassian instance is in the EU, US, or another supported region, Krina data stays in that same region automatically.
  • Programmatically verified by Atlassian. The badge is not self-declared. Atlassian's automated systems continuously check that the app meets the requirements. It is removed automatically if compliance lapses.
<200ms

Cached response

Once classified, your inbox loads in under 200ms. New @mentions are classified in under 2 seconds — before you've finished reading the thread.

— Mark, Gonzalez & Harris · ACM CHI 2005

Privacy by

Architecture — GDPR Art. 25

Only the comment you wrote is analysed — no emails, no credentials, no browsing data. Display names in @mentions are included for classification accuracy.

— Cavoukian · Privacy by Design, 2009

90 days

Auto data eviction (default)

Action items are automatically cleared after 90 days by default.

Each user can configure their own retention window — from 30 days up to 365 — in Settings.

— Iqbal & Bailey · ACM CHI 2006

The Product

Priority dashboard and action inbox — every @mention across Jira, Confluence, and Jira Service Management — classified, grouped by urgency, and ready to act on.

Features

Everything you need. Nothing you don't.

3D AI Classification

Every @mention classified across three independent dimensions simultaneously: Action (RESPOND, REVIEW, APPROVE, FIX, BUILD, WATCH, NOTE), Urgency (CRITICAL, HIGH, NORMAL, LOW with timeframe context), and Impact (CUSTOMER, REVENUE, TEAM, INDIVIDUAL). Not a single score — a full picture.

In Detail

Action, Urgency, and Impact — visible at a glance.

Every @mention gets a three-dimensional classification — so you know exactly what to do, how fast, and who it affects.

Why three dimensions? Urgency and Impact map directly to the Eisenhower Matrix — the prioritisation framework engineering teams already know. The Action dimension is the layer the Eisenhower Matrix never had: RESPOND · REVIEW · APPROVE · FIX · BUILD · WATCH · NOTE. It answers not just when to act, but exactly what kind of action is required. This combination does not exist natively in Jira or anywhere else in the Atlassian ecosystem.

Urgency × Impact: Eisenhower Matrix · Action dimension: Bellotti et al., ACM CHI 2003

Inline tooltips explain why each item was classified — full context without leaving your inbox.

Personal Priority Inbox

All @mentions from Jira, Jira Service Management, and Confluence in one unified inbox. Critical items surfaced first. Unread tracking, resolve with 10-second undo window, and bulk resolve for processing multiple items at once. Items with approaching deadlines surface a live countdown badge automatically — no manual flagging required.

Follow-ups

Mark any @mention with a due date and track it in a dedicated Follow-ups view, sorted by date — so items you need to revisit don't get buried. Right-click any card and select Follow up… to set a date, or click the calendar icon directly.

Dashboard to Classification

Drill from a high-level dashboard view straight into the underlying classified items. Click any priority, urgency, or impact bucket to see exactly which @mentions sit behind it — context preserved, filters auto-applied.

Admin Configuration

Jira admins control which Jira projects, Jira Service Management projects, and Confluence spaces are monitored. Include or exclude mode per product. Changes apply immediately across all users — no per-user configuration required.

Advanced Filtering & Search

Filter by Action type, Urgency, Impact, Project, Issue Type, Author, Time Range, Read Status, and Source. Full-text search. Stackable filter chips. All filters persist per user session.

Grouped Views

Switch between flat list, grouped by issue, or grouped by source (Jira / JSM / Confluence). Collapsible groups. View mode and sort order persisted per user.

Smart Auto-Refresh

Your inbox stays current automatically. A visual refresh indicator shows sync status at a glance, with manual refresh always available.

Deadline-Aware Urgency

Deadline phrases in @mentions — EOD, COB, ASAP, "end of sprint", "before the release", "by Friday" and 40+ more — are detected automatically. Items approaching their deadline escalate urgency and show a live countdown ("⚠ Due in 3 days") without you having to flag anything manually.

Mute Controls

Mute noisy projects or specific authors. Toggle muted item visibility without losing coverage. Cut through the noise while keeping full audit trail.

Storage Management

Monitor your inbox storage at a glance. Clear resolved items, older items, or reset entirely — all from within your user settings. Data older than 90 days is automatically evicted by default.

Grouped by Issue

When a hot ticket explodes, nothing gets missed.

A single Jira issue, Confluence page, or JSM ticket can rack up dozens of @mentions in minutes. Krina collapses every mention on that one item into a single thread — sorted by urgency, with CRITICAL items surfaced first — so the right people respond immediately without scrolling through comment history.

Group by issue — every @mention on a single Jira issue, Confluence page, or JSM ticket collapsed into one thread, sorted by urgency. CRITICAL items always surface first.

The AI Engine

Classification
at machine speed.

Built on Atlassian Forge with multi-layer PII sanitisation and near-instant classification.

New @mentions classified in under 2 seconds; cached responses under 200ms
  1. 1

    @mention detected

    Forge trigger fires on Jira, Jira Service Management, or Confluence comment events. Zero mentions? Processing stops immediately — nothing stored.

  2. 2

    Multi-layer PII sanitisation

    Emails → [EMAIL]. Phone numbers → [PHONE]. API keys & tokens → [SECRET_N]. Account IDs → [USER_N]. Customer names → [CUSTOMER_N]. HR/legal terms → [SENSITIVE]. Credit cards → [CREDIT_CARD]. Personal name signatures → [NAME]. Sanitised text only proceeds to AI.

  3. 3

    AI classification

    Sanitised text and basic issue context (title, type, priority, status) are processed by Atlassian Forge AI — inference runs inside Atlassian's own infrastructure, never on external servers. The response is validated and post-processed — including automatic deadline extraction from 40+ phrase patterns (EOD, ASAP, end of sprint, before the release) and urgency escalation for items approaching their due date.

  4. 4

    Near-instant delivery

    New @mentions are classified in under 2 seconds. Cached responses load in under 200ms.

  5. 5

    Resolve, mute, or act

    Resolve in one click, mute noisy authors, or jump straight to the source comment in Jira, JSM, or Confluence.

Intelligence Layer

See your workload differently.

The Urgency × Impact matrix gives you an instant read on where pressure is building across your team — before it becomes a crisis. Click any cell to filter your inbox to exactly those items.

Matrix view — Urgency × Impact. Click any cell to instantly filter your inbox to that category.
Privacy & Data Handling

Your data never leaves Atlassian infrastructure.

Krina is built entirely on Atlassian Forge — Atlassian's own serverless platform. All data is stored exclusively in Forge SQL — Atlassian-hosted relational database. There is no external database. There are no external servers. AI classification is performed by Atlassian Forge AI, which runs inside Atlassian's own infrastructure — sanitised text never leaves the Atlassian platform. Emails, phone numbers, API keys, and other sensitive values are replaced with typed tokens before the AI call. Atlassian display names from @mentions are passed to the AI for classification accuracy.

Action items (your @mentions)Stored: Forge SQL — Atlassian-hosted relational databaseRaw PII: No — account IDs anonymised
AI classification resultsStored: Forge SQL — Atlassian-hosted relational databaseRaw PII: No — contains only AI output, never comment text
AI classification cacheStored: Forge SQL — Atlassian-hosted relational databaseRaw PII: No — contains only AI output, never comment text
User preferences & settingsStored: Forge SQL — Atlassian-hosted relational databaseRaw PII: No
Admin configurationStored: Forge SQL — Atlassian-hosted relational databaseRaw PII: No
AI inference inputStored: Atlassian Forge AI (in-platform), not storedRaw PII: No — multi-layer PII sanitisation applied first
Scopes

Permissions requested — and why.

Krina requests only the permissions it actively uses. Here is every permission and exactly why it is needed.

read:jira-workRead Jira issue details, comments, and attachments when an @mention is detected
read:jira-userRead author display names for inbox rendering
storage:appRead and write to Forge SQL (your personal inbox data)
read:confluence-content.summaryDetect Confluence comment events
read:content:confluenceRead Confluence comment body text
read:comment:confluenceRead Confluence comment metadata
read:space:confluenceRead Confluence space information for admin configuration
read:page:confluenceRead Confluence page context for comment display
read:user:confluenceRead Confluence author display names
read:servicedesk:jira-service-managementAuto-detect JSM projects for admin configuration
No external (non-Atlassian) network calls. AI inference runs on Atlassian Forge AI, inside the platform.

On the roadmap

More products in development.

Krina is one of several AI-native products we're building for engineering and operations teams. Follow us or get in touch to hear about what's next.

Built on Evidence

Every design decision maps to published research.

Krina was not designed by guesswork. The problem, the classification model, the privacy architecture, and the AI pipeline each map to peer-reviewed research in human-computer interaction, cognitive science, and AI.

THE PROBLEM

In 1996, researchers identified that knowledge workers were using email as an accidental task manager — a communication tool pressed into service as a to-do list. In 2026, your team is doing the same thing with Jira comments.

Whittaker & Sidner · ACM CHI 1996 Bellotti et al. · ACM CHI 2003

THE COST

Knowledge workers average 3 minutes on a task before context-switching, across 10 working spheres per day. Interruptions don't just cost time — they degrade cognitive state, increasing stress and error rate. The inbox is the resumption aid this research implied was needed.

Mark, Gonzalez & Harris · ACM CHI 2005 Mark, Gudith & Klocke · ACM CHI 2008

THE DESIGN DECISION

Interruption cost is not constant — it depends on where in a task the interruption lands. Interruptions at subtask boundaries cost significantly less than mid-task interruptions. Krina batches new items on a set interval rather than pushing notifications in real-time. New items surface between tasks, not during them.

Iqbal & Bailey · ACM CHI 2006

THE FRAMEWORK

Urgency and Impact map directly to the Eisenhower Matrix — the most cited prioritisation framework in productivity research. The Action dimension (RESPOND · REVIEW · APPROVE · FIX · BUILD · WATCH · NOTE) is the novel extension: it answers what kind of action is required, not just when. This combination does not appear in prior published literature.

Eisenhower Matrix Bellotti et al. · ACM CHI 2003

PRIVACY

The 8-layer PII sanitisation pipeline is structural, not configurable. It runs before every AI call. This is Privacy by Design Principle 1 — Proactive not Reactive — now codified in GDPR Article 25. Sensitive values are stripped before every AI call. Atlassian display names in @mentions are included for classification accuracy — this is a deliberate design decision.

Cavoukian · Privacy by Design, 2009 GDPR Article 25

THE AI

Classification uses chain-of-thought prompting with cross-industry few-shot examples — industry-neutral, covering healthcare, finance, legal, HR, engineering, and sales. Zero-shot generalisation means accurate classification across any team's Jira instance without fine-tuning.

Wei et al. · arXiv 2022 Sanh et al. · ICLR 2022 Ouyang et al. · NeurIPS 2022

The researchers and institutions cited above have not reviewed or endorsed this product. Citations reflect published findings that are consistent with the design decisions made independently in building Krina.

Stop losing decisions in comment threads.

Krina is live on the Atlassian Marketplace. Install it free and give every team member a personal, AI-classified inbox for every @mention across Jira, Confluence, and Jira Service Management.