Vector
Published Sep 6, 2026Issue tracking that stays fast and stays yours.
Replaces
- Jira Software · Atlassian Corporation — $18.30 per seat per month (Jira Cloud Premium (rate at 100 users)), about $87,840/yr at 400 seats
- Linear · Linear Orbit, Inc. — $16 per seat per month (Business (billed yearly)), about $76,800/yr at 400 seats
Rented, per year
$87,840
Jira Software list price at 400 seats.
Owned, per year
$8,280
AWS infrastructure only. Azure is $8,940.
Build
8 weeks
2 engineers. To parity on the ten core features above, including migration of existing Jira projects with issue keys, comments, attachments and sprint history preserved.
Difficulty
3/5
Well-understood shape. Low delivery risk.
Reference organisation: 400 engineers, 800 total users, ~1.5M issues, 40 GB attachments, single region, prod + non-prod. 3-year Compute Savings Plan on Fargate, 3-year reservation on Container Apps and PostgreSQL Flexible Server. Non-prod stopped outside business hours. Excludes vendor support plans, LLM gateway token spend and egress above 1 TB/month. Figures are infrastructure only and exclude the build.
What it is
Vector is issue tracking for engineering teams: issues, cycles, projects, triage, keyboard-first navigation and a query language, running on a Postgres database inside your own account. It targets engineering organisations of 200 to 2,000 developers, where Jira has accreted a decade of custom fields and workflow schemes that nobody can safely delete, and where the annual renewal now costs more than a senior engineer. Vector keeps the parts teams use every day and drops the configuration surface that produced the sprawl.
Why this one stops making sense
- At 800 licensed users — Jira has no free read-only seat, so product and leadership pay seller rates to read a backlog — Jira Cloud Premium is roughly $175,700 a year at the published 100-user rate, and Linear Business is $153,600. Neither leaves you with an asset.
- Your issue history is the record of how your engineering organisation actually works. It lives in someone else's schema behind a rate-limited REST API with a 10,000-result search ceiling.
- Jira's configuration model — schemes, contexts, screens, custom fields shared across projects — makes cleanup risky, so nobody cleans up, so it gets slower every year.
- Automation rules, dashboards and JQL filters accumulate as untracked production logic with no code review and no rollback.
- Marketplace apps that started as a $2/user convenience become load-bearing, and each one is a second renewal negotiation you do not control.
What ships
- Issues with parent/sub-issue nesting, estimates, labels and relations
- Cycles and a triage inbox with automatic carry-over of unfinished work
- Projects and initiatives that roll up progress from issue state, not manual status
- A typed query language over issues with saved, shareable views
- Keyboard-first navigation: every action reachable without the mouse
- Workflow states defined per team, versioned in git alongside the schema
- Git integration: branch naming, PR linking and automatic state transition on merge
- Real-time updates over WebSocket with optimistic local writes
- Full-text and trigram search across issues, comments and attachments
- Cycle analytics: scope change, throughput, and time-in-state distributions
Data model
Stack
- frontend
- Next.js 15 App Router + Tailwind + TanStack Query, with a local IndexedDB mirror for optimistic writes
- backend
- NestJS on Node 22 — REST plus a WebSocket gateway for issue deltas
- database
- PostgreSQL 16 with Row Level Security per team and one read replica
- cache
- Redis 7 for presence, view result caching and read-your-writes routing
- queue
- BullMQ for git webhook processing, notification fan-out and analytics rollups
- auth
- OIDC against the customer IdP; SCIM 2.0 for team membership
- search
- PostgreSQL tsvector plus pg_trgm for fuzzy title matching
- ai
- Optional duplicate detection and issue summarisation through a pluggable LLM gateway
Parity, honestly
What we match, and what we do not.
The rows marked No are the important ones. If one of them is the reason you bought Jira Software, keep buying it — and we will tell you that on the first call rather than the fifth month.
| Capability | Jira Software | Ours | Note |
|---|---|---|---|
| Issues, sub-issues, relations and dependencies | Yes | Yes | — |
| Sprints / cycles with velocity and burndown | Yes | Yes | — |
| Query language over issues (JQL equivalent) | Yes | Yes | Vector's parser compiles to parameterised SQL, so a saved view is a plan you can EXPLAIN. |
| Keyboard-first interaction model | Partial | Yes | Linear sets the bar here; Jira does not. Vector matches Linear. |
| Git provider integration (GitHub, GitLab, Bitbucket) | Yes | Yes | Webhook-driven, plus a CLI for local branch creation. |
| Free unlimited viewer seats for non-engineers | No | Yes | Product, support and leadership read the backlog without a licence. |
| Real-time multiplayer updates | Yes | Yes | — |
| Roadmaps and initiative roll-up | Yes | Yes | — |
| Native iOS and Android apps in the app stores | Yes | Partial | Vector ships an installable PWA with offline read. Store-listed native apps are a separate 4-week engagement. |
| Atlassian Marketplace (5,000+ apps: Tempo, Structure, Xray, ScriptRunner) | Yes | No | We do not rebuild a marketplace. If you depend on Xray for test management or Tempo for timesheets, either keep them or scope them as their own build. |
| Vendor-run SOC 2 Type II and ISO 27001 attestation | Yes | No | Vector inherits your own cloud compliance posture. If your customers require a third-party report naming the software vendor, that is a reason to keep buying. |
| Jira Data Center / on-premises parity for air-gapped sites | Yes | Partial | Vector runs anywhere Postgres and containers run, including air-gapped, but we do not ship an installer with a support matrix. |
| AI issue summarisation and duplicate detection | Partial | Yes | Runs through your own LLM gateway. Nothing is used for training. |
| Migration from existing Jira projects | None | Yes | Issues, comments, attachments, custom fields, sprint history and issue keys are preserved so old links resolve. |
Reference architecture
Built twice, on purpose.
Every application in Techtons ships with a production architecture for AWS and one for Azure, using the same diagram grammar so you can read them side by side. Your cloud, your account, your bill.
Amazon Web Services
- Compute
- ECS Fargate services (api, ws, worker) behind an ALB
- Data
- Aurora PostgreSQL with one reader instance, ElastiCache Redis
- Storage and edge
- S3 for attachments, CloudFront for static output and file delivery
- Identity
- Cognito federated to the customer IdP via OIDC/SAML
- Observability
- OpenTelemetry to CloudWatch and X-Ray, RUM on the front end
- Infrastructure as code
- Terraform, one module per environment, remote state in S3 with a DynamoDB lock
Services
The decision worth arguing about
The non-obvious decision is refusing to add OpenSearch. Every issue tracker eventually grows a search cluster, and it becomes the most expensive and most fragile component in the system — a second copy of your data, a second consistency model, and a nightly reindex that fails silently. Vector runs search as tsvector plus pg_trgm on the Aurora reader instead. The trade is real: you lose BM25-grade relevance ranking and cross-project aggregation over a hundred million documents, so a 5,000-issue result set orders by recency and match count rather than by learned relevance. In exchange there is one stateful system, replica lag sits under 100ms so an issue is searchable before the create toast fades, and the search index cannot drift from the source of truth because it is the source of truth.
Microsoft Azure
- Compute
- Azure Container Apps (api, ws, worker) with KEDA scaling on queue depth and HTTP concurrency
- Data
- Azure Database for PostgreSQL Flexible Server with a read replica, Azure Cache for Redis
- Storage and edge
- Blob Storage for attachments, Front Door Standard for edge and WAF
- Identity
- Microsoft Entra ID directly, no broker layer
- Observability
- Azure Monitor with Application Insights, OpenTelemetry-native
- Infrastructure as code
- Bicep in an azd-compatible layout
Services
The decision worth arguing about
Azure differs from AWS in one place that matters for a keyboard-driven tool. Flexible Server read replicas are asynchronous with no equivalent of Aurora's session-consistency endpoint, and Vector's core interaction is create-an-issue-then-immediately-filter, which lands on the replica milliseconds after the write. The fix is a read-your-writes gate: after any mutation the user's ID is written to Redis with a 3-second TTL, and every request from that user during the window is pinned to the primary. The trade is that a burst of writers pushes read traffic back onto the primary exactly when it is busiest, so the primary is sized one SKU above what steady-state reads would suggest — roughly $60 a month more than the AWS shape for the same load.
Sources
Every price on this page, with the page we read it from and the date we read it.
| Product | Plan | List price | Checked | Source |
|---|---|---|---|---|
| Jira Software | Jira Cloud Premium (rate at 100 users) | $18.30 / seat/month | 2026-09-05 | community.atlassian.com/forums/App-Central-a Atlassian prices Jira Cloud progressively rather than at a flat per-seat rate, so there is no single list number; $18.30 is the published Premium rate at 100 users as of August 2026 and falls as seat count rises. |
| Linear | Business (billed yearly) | $16 / seat/month | 2026-09-05 | linear.app/pricing |
Should you still be paying for Jira Software?
Two weeks, fixed price. We audit what you actually use, map it against Vector, cost the replacement on your own AWS or Azure account, and give you a delivery plan. If the honest answer is to keep the licence, that is what the report will say.