Defensible research fortax & legal professionals
Autherise Scribe researches regulations across 10+ jurisdictions, analyzes your case files, and drafts cited, export-ready artifacts — letters, appeals, memos — in minutes, not days.
Based on Directive 2006/112/EC Art. 273 and CJEU ruling C-324/20, the assessment appears procedurally defective on three grounds:
- Failure to provide a pre-assessment hearing.
- Misapplication of the reverse-charge mechanism.
- Statute of limitations under national law.
- Identify grounds
- Gather precedent
- Draft letter
- Verify citations
Built for the work you actually do.
Autherise Scribe is an AI research assistant engineered for rigor — not a chatbot wearing a legal costume.
Plan-then-execute
Complex requests produce an approvable plan. You stay in control — the backend executes it step by step.
Multi-jurisdiction research
Cascading search across Brave, Exa, and Tavily with round-by-round synthesis. Every fact traces to a source.
Living artifacts
Open a document, chat about it. Surgical edits, section rewrites, inline diffs — all versioned and reversible.
Cited & defensible
Inline citations on every claim. Export letters, appeals, memos with your branding — and the audit trail intact.
Refresh-resilient
AI runs server-side on Convex. Close the tab mid-research — it continues. Come back, pick up exactly where you left off.
Fast where it matters
Opus for drafting. Sonnet for chat. Haiku for extraction. Prompt caching keeps latency low and costs predictable.
Documents that evolve with the conversation.
Every artifact is a structured object the AI can edit surgically. Versioned, diffed, reversible — production-grade writing instead of brittle generation.
Dear Commissioner,
We write on behalf of our client to formally object to the above-referenced assessment on three procedural grounds, each sufficient in itself to vacate the determination.
First, the assessment was issued without the pre-hearing required under Directive 2006/112/EC Art. 273, as clarified by the Court of Justice in C-324/20…
…
From question to defensible artifact.
Three steps. No black box.
Ask — or upload
Start a chat. Upload case files. Scribe extracts entities, builds summaries, and pins the relevant jurisdictions.
Review the plan
For complex work, Scribe drafts a plan. Approve it, edit it, or just let it execute. You always see what it will do before it does.
Get a cited artifact
Letters, appeals, memos, presentations, infographics — every claim cited, every export branded, every version reversible.
Simple, credit-based, predictable.
Every plan includes cited exports, refresh-resilient sessions, and version history.
- 5 credits / month
- 1 jurisdiction
- 1 deep-research session / month
- PDF export (watermarked)
- Community support
- 50 credits / month
- 4 jurisdictions
- Unlimited deep research
- All export formats — no watermark
- Claude Opus 4.6 on everything
- Priority support
- 200 credits / month
- All jurisdictions
- Unlimited deep research
- All formats + API access
- Team workspaces
- SSO & audit logs
“Autherise Scribe compressed three days of cross-jurisdiction research into forty minutes — and every citation held up under review. It's the first AI tool I trust with work the partners will sign.”
Answers to the things you were going to ask.
How does Scribe keep work from disappearing if I close the tab?
Every AI operation runs server-side on Convex. Progress, partial content, plan approvals, and streaming output all live in the database — so refreshing or closing the tab never loses state. When you come back, you see exactly what the server sees.
Where do the citations come from?
Scribe cascades across Brave, Exa, and Tavily, then synthesizes findings round-by-round. Every claim in a generated artifact links to the exact source it came from. If a source can't be verified, the claim is flagged — not smoothed over.
Can I edit artifacts surgically, or only regenerate?
Both — but editing is the default. You can tell Scribe to tighten a section, insert a paragraph after a specific citation, or swap a date, and it produces a diff you accept or reject. Full regeneration is reserved for when you actually want it.
Is my data used to train models?
No. Uploads, chats, and artifacts are stored in your workspace and not used for model training. Team plans include audit logs and SSO for compliance-sensitive workflows.
What's a credit, and how many does a typical task use?
Credits are consumed per billable action — a deep-research session, a long artifact generation, or a premium export. Short chat turns and light edits are free. Your dashboard shows exactly what each action will cost before you run it.
Stop hunting citations. Start shipping defensible work.
Five free credits. No card required. Your first letter could be drafted in the next ten minutes.