Control you can prove.
Simplicity your team will actually use.
Most AI platforms make you choose between governance and usability. Romona is built so you don’t have to.
Self-hosted & governed
Romona runs on your infrastructure: cloud, on-premise, or fully air-gapped. Your documents never leave your perimeter, nothing trains an external model, and every AI call is logged with a grounding score, cost and full trace. Sovereignty and assurance, not just a dashboard.
- No data leaves your environment, ever
- Grounding & hallucination score on every answer
- Immutable, exportable audit trail per call
- RBAC, SSO and data-retention controls built in
Built for business users
No prompt engineering required. Pick an agent, ask in plain language, and get answers grounded in your company’s knowledge, with citations you can click. Admins compose agents, skills and workflows through guided, no-code builders. Everyone else just gets work done.
- Guided, no-code agent & workflow builders
- Plain-language chat with clickable citations
- Reusable skills for consistent, on-brand output
- Modes for Analyse, Research, Design & Build
Everything your organisation needs.
Nothing it doesn’t.
One coherent system for knowledge, agents, automation and governance. Not a pile of bolt-ons.
Grounded in your own knowledge
Upload documents or connect live sources. Romona indexes everything so every answer is grounded in your data, with a citation on every claim.
- →Any source: PDFs, spreadsheets, docs, or live connectors for GitHub, Jira, Confluence, Notion, Drive, SharePoint and Salesforce.
- →Cited by default: every response links back to the exact source it used.
- →Scoped workspaces: per-client and per-team boundaries, with no leakage between them.
Governed agents, built from your stack
Compose specialists from your knowledge, skills and tools, then publish them org-wide. Each one is a governed expert, not a generic chatbot.
- →Purpose-built: a Legal Reviewer, Sales Analyst or DevOps Assistant, scoped to the right KBs.
- →Reusable skills: encode best practices for consistent tone, terminology and format.
- →Tools & MCP: connect internal APIs by JSON schema, or plug in MCP servers for a whole ecosystem.
- →Connectors: agents stay current by syncing the tools your team already runs on.
Connects to the tools your team already uses
One platform, the right mode for the task
Romona adapts to the work in front of your team, giving each task the right tools and defaults instead of one blank chat box.
- →Analyse & Deep Research: grounded insight, and multi-step research that writes its own cited report.
- →Designer: describe UI, decks or campaigns in plain English and preview them rendered live.
- →Build: functional apps, scripts and data pipelines, executed in a secure sandbox.
From one-off chats to repeatable processes
Turn the same governed agents into automation. Design pipelines on a visual canvas and let specialist agents collaborate, every step logged.
- →Visual workflow builder: drag-and-drop agents, tools and conditions. No code.
- →Multi-agent orchestration: LangGraph graphs run sub-agents in parallel, trace streamed live.
- →On demand or scheduled: with a full execution trace logged per run.
Oversight on every interaction
Governance is the default, not an add-on. Admins decide who can use what, and every call is logged with a grounding score and full trace.
- →Controls built in: RBAC, audit logs, data retention and SSO across the whole platform.
- →Every call traceable: grounding score, cost, tool calls and sources, exportable to your SIEM.
- →Your model, your rules: Claude, OpenAI, Azure, Bedrock or air-gapped Ollama, with per-model cost caps.
Built for organisations where
data governance is non-negotiable.
Security was a first principle here, not bolted on afterwards. Every feature is built with enterprise compliance in mind.
Zero Data Retention
Ephemeral mode persists nothing: responses stream straight to the user. Your data never leaves your infrastructure, and never trains a model.
Role-Based Access Control
Granular per-resource permissions, custom roles and project-level membership. Regular members can’t reach administrative functions.
SSO & SAML 2.0
Azure AD, Okta or Google Workspace via SAML/OIDC. Disable the company account and Romona access is revoked automatically.
Immutable Audit Logs
Who asked, which model ran, sources cited, grounding score, tool calls and full trace, logged per call and exportable to your SIEM.
Multi-Tenant Isolation
Row-level security on every query, with org_id on every record. Defence-in-depth across API, middleware and database.
Data Residency
Deploy in your region: EU, US or on-premise. Built for GDPR, SOC 2 Type II and ISO 27001 compliance postures.
SSRF Protection
Every outbound tool URL is validated before the request leaves. Private IP ranges and metadata endpoints are blocked by default.
Rate Limiting & Account Security
Logins are rate-limited and locked after repeated failures. Short-lived access tokens; refresh tokens hashed at rest.
Prompt Injection Defence
User input is delimited before reaching the model, separating instructions from data. Security headers block XSS and clickjacking.
Your infrastructure. Your rules.
Run Romona wherever your organisation needs it: managed cloud, self-hosted, or fully air-gapped. The full feature set everywhere, with zero external data calls, deployed by one engineer and in use across the org within the week.
Railway / Managed Cloud
One-click deploy to Railway with auto-migration, object storage (Cloudflare R2), and managed PostgreSQL. Up in under 10 minutes, zero ops overhead.
- Auto-scales with traffic
- Zero-downtime deploys
- Managed SSL and domains
- Built-in secrets management
- Bring your own cloud account
Docker Compose
Full local or on-premise stack in a single docker compose up. All services included: Postgres, Redis, MinIO, Nginx, Celery.
- Complete data sovereignty
- Hot reload for customisation
- Custom networking and VPN
- Internal CA / TLS supported
Kubernetes / Air-Gapped
Production-grade Helm charts for enterprise teams, or bundle all images as a tarball for completely isolated networks with on-prem LLM models via Ollama.
- AWS, GCP, Azure, or bare metal
- Horizontal autoscaling
- Ollama for air-gapped LLM inference
- Keycloak for SSO / SAML
- No internet required at runtime
Runs on
LLM providers
Data & integrations
Ready to govern AI
at your organisation?
Request a demo and we’ll show you Romona running on your infrastructure within the week.
Self-hosted · No data sharing · Enterprise pricing available