Your Linear issues, connected to the code that ships them .
Oracus.ai reads your Linear workspace and links every issue to the services it affects, the team that owns it, and the customer flow it touches. So a product owner, campaign manager, or customer lead can ask what shipped, what's blocked, and what it affects — with citations, without flagging an engineer.
What your team can finally ask about Linear
The questions that live in a Linear board but historically needed an engineer to answer — because the answer is only obvious once issues are joined to the code, services, and owners behind them. Every answer cites its sources.
› Which Linear issues are blocking the TikTok v3 migration — and who owns them?
Four issues across two teams. Channels owns the sync rewrite (ENG-2847, in progress); Analytics owns the metric-name remap (ENG-2903, in review). Both are sub-issues of the v3 epic. Tech leads on the hook: @miguel and @ana.
› Did we actually ship everything in the 'Cross-channel preview' project?
Partially. The project's spec covers all four channels — today only Meta and X render a real preview, while TikTok and LinkedIn still return a placeholder. Two issues in the project are still open and unassigned.
› What's the blast radius of ENG-2847 before we start it?
It rewrites tiktok-sync, which feeds reporting-aggregator and billing-engine downstream. In production terms: every active ad-group targeting TikTok — 132 campaigns across 47 advertisers as of the last sync. Worth a heads-up to the Analytics and Billing owners first.
› Which bugs in the current cycle touch billing?
Two. GROW-512 (invoice rounding on multi-currency advertisers, in progress) and GROW-548 (double-charge on retried campaigns, in review). Both are labelled for billing-engine; neither has a linked test yet, so coverage is thin.
› What is @ana's team carrying that touches creative validation?
Three in-progress issues on creative-validator: TikTok MP4 spec support (ENG-2871), a LinkedIn aspect-ratio rule (ENG-2884), and a validation-timeout bug (ENG-2890). The first two are sub-issues of the channel-parity epic.
› Any Linear issues sitting on channels nobody has deployed to in 90 days?
Yes — three open issues target linkedin-sync and x-sync, neither of which has seen a deploy or PR since Q4. Both integrations had upstream API breaking-change notices in the meantime, so those issues are likely aging into risk.
What the connector wires up
Nothing to configure issue-by-issue. Oracus.ai reads what's already in Linear — the hierarchy, the labels, the workflow — and joins it to your code so the answers hold together.
Issues → services & code
Each Linear issue is linked to the services and repos it touches, so 'what does this issue affect?' has a real answer — not a guess. Uses your labels; no manual tagging in Oracus.
Sub-issues → epics & projects
Linear's parent/sub-issue hierarchy comes across intact, so a question about an epic rolls up every child issue and the code beneath it.
Labels & workflow state
A 'Bug' or 'Story' label becomes the issue type; the workflow state becomes status, and a done/cancelled state records the outcome — so 'what's in review that touches billing?' just works.
One graph, every source
Linear lands in the same knowledge base as your code, tests, and specs — and alongside Jira if you run both. Cross-source questions ('spec says X, did we ship it?') answer in one place.
How the sync works
Read-only, keys stay yours, self-hosted by design. Oracus.ai pulls from Linear on a schedule so the picture never drifts far from reality — and alongside Jira, if your teams use both.
- 1
Connect your workspace
A read-only Linear API key is all it takes. Oracus.ai pulls issues, sub-issues, labels, and team metadata — nothing is written back to Linear.
- 2
Links every issue to reality
Each issue is connected to the services that implement it, the team that ships it, and the customer flow it touches. The graph is what makes cross-cutting questions answerable.
- 3
Answers with citations
Every claim names the Linear issue, service, or owner it came from. The prompt enforces it; the response schema requires it. Hallucinations are structurally hard.
Point Oracus.ai at your Linear
We'll connect a workspace and a handful of repos, then walk you through the questions your board can suddenly answer — no install on your side, no commitment.