DOORS holds the record.Pellucid reads it.
DOORS has held the canonical record for two decades of defence and aerospace programmes. It is, by design, a database of requirements — not a reader of them. Pellucid sits alongside DOORS as a reading layer for the prose inside each module.
Where they overlap.
Where they differ.
Two tools, two jobs. IBM DOORS keeps the record. Pellucid reads the prose. Most regulated programmes need both.
DOORS has long supported a small set of attribute-based linting — vague-term searches, custom DXL scripts that count weak modals, the occasional bespoke regex. Many programmes have built versions of this themselves. Pellucid covers the same surface on its rule layer in milliseconds, with a much larger INCOSE-aligned lexicon. Past that first pass, the two tools diverge entirely.
DOORS keeps the artefact honest across decades. Pellucid keeps the sentence honest before the artefact ships.
DOORS is a database with a UI on top — it is excellent at being the long-lived record of who said what, when, and which baseline they said it in. It was never going to be a reader of natural language; it doesn't need to be. Pellucid is the reader. After the rule layer, four specialists run in parallel through LangGraph: a lexical agent, a syntactic agent, a domain agent calibrated on your baselines, and a risk agent. Each proposes findings; each critiques the others. The aggregator keeps every vote on the record and produces a single calibrated score. The findings then write back to DOORS as object attributes, so the work of reading lives in the same place as the work of recording.
- 01
A formal object model with rigorous baselines, modules, links, and a long lineage of audit-grade traceability.
- 02
DXL scripting and the OSLC interface that let mature programmes attach custom analysis at the object level.
- 03
Established conventions for level-of-requirement separation (stakeholder, system, sub-system, component) that downstream V-model tooling relies on.
- 04
Proven scaling on long-lived defence and aerospace programmes where decade-spanning history is required.
- 05
A familiar surface for programmes governed by MIL-STD-498, DO-178C supplements, and CENELEC EN 50128.
- 01
A reading layer that adds INCOSE-aligned ambiguity detection — vague modals, passive constructions, unbounded quantifiers — far past the small built-in lexicon.
- 02
A four-specialist panel (lexical, syntactic, domain, risk) that debates each finding and emits a calibrated confidence score.
- 03
Attribute write-back via OSLC or DXL: every Pellucid finding lands as a custom attribute on the DOORS object, queryable in your existing views.
- 04
Domain calibration trained on your historical baselines — so the panel reads the language your programme actually uses, not generic prose.
- 05
Audit-friendly export of the panel's reasoning chain, alongside the DOORS history record, for regulator-ready traceability of *why* a clause was rewritten.
A real ambiguity Pellucid catches. Hover to read its mind.
The kind of sentence that sails through a review centre and detonates in design. Five readings, five products. Every span below ships with the panel's reasoning chain.
The flight-control software shall be robust and shall respond appropriately to unexpected sensor inputs within a reasonable timeframe.
Four ambiguities in nineteen words, in a system where ambiguity is a regulator-visible defect. DOORS will store this perfectly. Pellucid will refuse to let it ship.
Add Pellucid alongside DOORS.
Keep DOORS as the system of record. Add a reading layer that runs over each module and writes findings back as object attributes you can query in your existing views.
Pellucid integrates with DOORS Classic via DXL and with DOORS Next via OSLC. Findings land as custom attributes on the original objects. No migration, no parallel record, no second source of truth.