Un collègue IA qui connaît vos documents et répond en quelques secondes.
Importez vos contrats, manuels et tableurs. Posez vos questions en langage courant : Studio répond à partir de vos propres données et indique toujours de quel document et de quelle page provient l'information. Les données sensibles sont protégées automatiquement.
Un espace de travail RAG d'entreprise avec anonymisation des PII intégrée.
Recherche hybride (BM25 + pgvector), agent NL-SQL en lecture seule, automatisations de workflows, multi-fournisseurs (OpenAI / Anthropic / Google / Ollama local, BYOK) et une chaîne d'audit Ed25519 avec horodatage RFC 3161. La stack Privacy v2 est incluse dans chaque offre.
Le casse-tête du quotidien
Les réponses qu'il vous faut sont noyées dans des contrats, des manuels et des tableurs, et les dénicher prend des heures. Posez plutôt votre question en langage courant : vous obtenez la réponse, accompagnée de sa source exacte. Voici ce que ça donne.
Interrogez vos documents. Obtenez une réponse sourcée.
Studio répond à partir de vos propres documents et indique toujours la source. Cette démo connaît trois documents d'exemple d'une entreprise fictive — posez-leur n'importe quelle question.
Studio en production : recherche hybride (BM25 + pgvector) sur votre base de connaissances, réponse générée par le LLM de votre choix (OpenAI / Anthropic / Google / Ollama local) avec citation du document et de la page ; anonymisation des données personnelles dès l'ingestion. Cette démo utilise des réponses scriptées portant sur un ensemble fixe de trois documents — sans aucun appel à un LLM.
Six capabilities on one AI brain
Everything below sits on the same Corpilus AI engine, so the answers are consistent whether they came from chat, RAG, DB Reader, or a workflow step.
Knowledge base + RAG
Ingest contracts, manuals, SOPs. Studio chunks, embeds, and retrieves them at chat time — answers always cite the source document and page.
DB Reader
Read-only SQL agent over your operational tables. Ask in natural language; Studio writes the query, runs it, and returns a typed answer.
Workflows + Agents
Multi-step automations chained together — read a document, call an API, generate a report. Manual, scheduled, webhook, or event triggers.
Learning + personalisation
Studio learns the way your team writes — tone, glossary, preferred response shape. The longer you use it the more it sounds like you.
AI analyses
Compare documents, extract structured facts, run sentiment / risk / topic analyses over batches. Outputs land in a sharable report.
Multi-provider
OpenAI, Anthropic, Google, or our local Ollama — pick per tier or per request. BYOK supported. Same redaction stack on every route.
Privacy bundled in
Every prompt redacted before it leaves the tenant
Studio inherits the full Privacy v2 stack — 9-language pattern corpus, context-aware NER, kill switch, audit log. You don't buy Privacy separately when you have Studio.
How the redaction works →How Studio works end-to-end
From an ingested document to a cited, redacted answer — every step runs through one consistent pipeline.
Ingest
Upload PDFs, Word, Excel, Markdown, web pages. Studio chunks them, generates embeddings, and indexes into pgvector. PII is redacted at ingest, not later.
Retrieve
User asks in natural language → hybrid search (BM25 + vector) finds the right passages from your knowledge base, or runs read-only SQL against your DB.
Reason
Chosen LLM (OpenAI / Anthropic / Google / local Ollama) receives the redacted context + your question. Multi-step agent calls tools if the workflow needs them.
Cite + audit
Answer comes back with source links + page numbers. Every prompt and response is signed into the audit chain (Ed25519 + RFC 3161 timestamp).
Bring your own LLM — or use ours
No vendor lock-in. Switch providers per tier, per workspace, or per request. The same redaction stack protects every route.
OpenAI
GPT-4o · 4o-mini · o1 · o1-mini
Anthropic
Claude Opus · Sonnet · Haiku
Gemini 2.0 Pro · Flash · 1.5
Local Ollama
Llama 3 · Qwen · Gemma · Mistral
One Studio, every department
Studio adapts to how each team writes and what data they touch. The same brain, different views.
Finance
Reconcile invoices against the GL, ask GPT-4 about VAT cases with redacted amounts, draft monthly board memos from your bookkeeping data.
Legal
Compare 2,000 contracts to a master template, extract liability clauses, summarise NDAs. Party names redacted at ingest; LLM only sees structure.
Sales
Chat with the CRM, draft proposals from your product KB, generate localised follow-ups in 9 languages. Client names anonymised on outbound.
HR / People ops
Performance reviews with redacted PII, policy chatbot for employees, structured analyses across exit interview transcripts.
Engineering / Ops
Read-only SQL against your DB in natural language, runbook synthesis from incidents, on-call summaries from PagerDuty + Sentry.
Marketing
Draft copy with brand-tone learning, generate landing variants in 9 languages, summarise customer feedback into themes.
Ce que comprend chaque offre Studio
Des garanties concrètes, pas des chiffres pour faire joli.
Where Studio lives
Three delivery surfaces, one tenant config. Pick what suits the team — they all hit the same Corpilus AI brain.
PWA dashboard
Where your team writes prompts. Auto-redacts before any upstream call.
Desktop app (AgentPro)
Native runtime that pairs with your tenant. Fully offline reasoning. Coming soon.
MCP server
Native in ChatGPT Desktop and Claude Desktop. Studio tools available natively.
Ready to put Studio to work?
From €49 / month with Privacy v2 bundled. Bundle with Shield for −15% off the combined monthly total.
Privacy v2 bundled · 9-language pattern corpus · No vendor lock-in