CrystalSpec vs ReqView

ReqView, built by Eccam, a small Czech software company, is a lightweight requirements-management tool with a refreshingly honest posture: projects are local JSON files versioned in Git or SVN, edited in a desktop app on Windows, macOS, or Linux. With no vendor cloud backend, it is offline-capable, air-gap friendly, and popular with small automotive, aerospace, medical, and industrial teams that want ReqIF exchange, traceability matrices, and ISO 29148 templates without enterprise pricing. Prices are public on reqview.com/pricing: PRO at €440 per user per year, TEAM at €600, as of mid-2026.

CrystalSpec is the opposite architecture at a similar price point: an AI spec workspace in the cloud, where specs are typed entities — flows with branching decision points, data models, roles, test cases, glossary terms — versioned as revisions with field-level diffs, and every AI edit is a proposal a human approves. One plan, $10 per seat per month. Both vendors publish real numbers, so the table below can do the arguing.

Collaboration is the fault line

ReqView's most-cited weakness is structural, not cosmetic: a document cannot be edited by two people at once, and there is no Undo for structural changes. When requirements live in files, the team coordinates around the files — who has it open, whose commit wins, which export went to the stakeholder. CrystalSpec is built around that exact friction: editing forks a draft revision locked to one active editor, while everyone else keeps reading the stable published version. Discussions with @mentions live outside the edit lock, and stakeholders get read-only share links that need no account — instead of another emailed export.

Two legitimate versioning philosophies

To be fair, ReqView's versioning story is coherent: plain JSON in Git means baselines are commits, history is your repo's history, and requirements can sit next to code. For engineer-only teams, that is a feature, not a compromise. CrystalSpec versions decisions rather than files: a publish mints a revision whose change summary is AI-drafted and whose changeset resolves to individual fields, and every version stays revertible with full lineage. The difference shows when non-engineers arrive — a product manager can read a "View changes" diff; reconstructing history from Git commits in a desktop app is a harder sell.

The AI and agent gap

As of mid-2026, ReqView has essentially no AI beyond an NLP Similarity column that suggests related requirements. CrystalSpec was built the other way around. The assistant drafts flows, data models, roles, and test cases in the form of structured proposals, pre-validated with appliability checks and settled row by row; a declined proposal is preserved, along with who turned it down. Point the inconsistency analyzer at a project, a flow, or one step: it grades what it finds — contradictions, gaps, glossary terms nothing references — and offers each finding back as a reviewable fix.

The agent story diverges the same way. ReqView's JSON files are machine-readable in principle, but the vendor ships no agent tooling. CrystalSpec treats agent access as a product surface: a hosted MCP server (Claude Code, Claude Desktop, and Cursor verified), a scoped GraphQL API, and HMAC-signed webhooks fired at publish time. An agent can open a flow, enumerate revisions, and read the diff between them — and a publish converts those changes into atomic tasks inside GitHub, Linear, or ClickUp, every task tied back to its revision. Honest counterpoint: ReqView TEAM's Jira integration and bidirectional Azure DevOps sync are a genuine reason to pick it if those trackers are yours.