Consulting | Operations | Solutions | Cloud · Azure · AWS
Solutions
Products built by Stitch Labs from what we see every day on real projects.
RootViel
Agent-driven cybersecurity
Detect before you are detected
RootViel combines AI agents to assess your organisation's security posture before an attacker does: it analyses generative AI, servers, binaries, source code and networks, and runs online pentesting with full traceability and cryptographically signed reports (Ed25519).
Six technical capabilities, one interface
Generative AI
Models trained on TTPs, CVEs and contextual offensive intelligence.
Servers
Continuous inventory, fingerprinting and insecure configurations (CIS, DISA STIG).
Binary analysis
AI-assisted disassembly, malware signatures and obfuscation detection.
Online pentest
Controlled execution with attack path reporting.
Networking
Topology mapping, segmentation and firewall rules.
Source code
SAST and semantic review across more than 40 languages.
How it works
Asset registration
Import IP ranges, domains, repositories and cloud infrastructure, with authorised-scope validation.
Distributed scanning
Ephemeral agents run the scans from several locations, within the maintenance windows you define.
AI correlation
Models map CVEs, configurations and network paths into attack graphs with their exploitation chains.
Reporting and remediation
Technical and executive evidence with CVSS 4.0 scores, proofs of concept and containment playbooks.
What is a penetration test (pentest)?
Security testing in which evaluators mimic real-world attacks in an attempt to identify ways to circumvent the security features of an application, system, or network.
A triple safeguard on every pentest
Signed authorisation
Every run requires an authorisation signed with Ed25519.
Explicit windows
Actions only run within the agreed execution windows.
Dual human confirmation
Two people confirm before any destructive operation, with automatic rollback.
- Regulatory compliance
- ISO 27001:2022SOC 2 Type IINIST CSFPCI DSS 4.0HIPAAGDPR (art. 32 & 35)LOPDGDDLGPDNIS 2CCPA / CPRADORAEU AI Act
- Time to go live
- Starter48–72 h
- PRO+1–2 weeks
- Enterprise3–9 weeks, with a CSM from day one
- Verifiable and integrable
Every finding is signed (Ed25519 + SHA-256) and timestamped under RFC 3161: anyone can verify a report without logging in.
Integrates with your SIEM/SOAR: Splunk, Microsoft Sentinel, CrowdStrike, Palo Alto.
RootNia
Agentic knowledge platform
Your document knowledge, orchestrated by agents
RootNia connects your internal and external sources, classifies every document by sector, and answers with the right language model, all coordinated by Neural Intelligence Agents that keep the cycle moving without manual intervention.
- sectors classified: insurance, legal, finance, healthcare, engineering and logistics
- 6+sectors classified: insurance, legal, finance, healthcare, engineering and logistics
- LLMs integrated: ChatGPT, Claude, Kimi and more
- 3+LLMs integrated: ChatGPT, Claude, Kimi and more
- document origins: internal and external
- 2+document origins: internal and external
What RootNia does
Document orchestrator
Connects OneDrive, SharePoint and databases, from internal and external sources, into a single intake flow.
Extractor and classifier
Identifies and classifies every document by sector (insurance, legal, finance, healthcare, engineering and logistics) with no manual rules.
RAG
Retrieves the right context from within your own documentation, so every answer is grounded in your sources.
LLM integration
Connects to market models (ChatGPT, Claude, Kimi and whatever comes next) without switching platforms when the model changes.
Six specialized engines, one platform
From raw scans to grounded answers, each module handles one part of the pipeline so nothing gets lost in between.
MIKA
Multi-modal Intelligent Knowledge Analysis
Reads and interprets text, tables, images and scanned pages together, turning mixed-format documents into a single structured knowledge layer.
MOSAIC
Multi-Observer Specialized Analysis with Intelligent Consensus
Runs several specialized models over the same content and reconciles their outputs into one consensus result, reducing errors and inconsistencies.
VERA
Verified Extraction with Region Anchoring
Anchors every extracted value to its exact location in the source document, so any answer can be traced back and verified in a click.
OCR
Optical Character Recognition
Converts scanned pages, PDFs and images into searchable, machine-readable text.
RAG
Retrieval-Augmented Generation
Retrieves the most relevant passages from your knowledge base before generating an answer, keeping every response grounded in your own documents.
NL-to-SQL
Natural Language to SQL
Translates plain-language questions into SQL queries, so anyone can query structured databases without writing code.
What is RAG (retrieval-augmented generation)?
Retrieval-Augmented Generation (RAG) is the process of optimizing the output of a large language model, so it references an authoritative knowledge base outside of its training data sources before generating a response.
A cycle, not a series of steps: the Neural Intelligence Agents
Orchestrator
Ingests the document from your sources and hands it to the classifier.
Classifier
Sorts it by sector and feeds the RAG layer.
RAG
Retrieves the context and picks the right LLM to answer.
Back to the orchestrator
The result flows back to the orchestrator: no manual queues between steps.
Built to operate inside your controls, not around them
RootNia is designed so every retrieval is logged, every document tier is access-controlled, and nothing bypasses the rules your organization already has in place.
Designed to help you comply with
- GDPR
- EU AI Act
- NIS2
- DORA
- LOPDGDD
Infrastructure aligned with
- ISO 27001
- SOC 2 Type II
- ENS
Governed retrieval
Every answer is traceable
Each response links back to the documents that grounded it. You can audit which source, which sector classification and which LLM produced any given answer.
Access by design
Private, public and technical tiers
Documents are separated into private, public and technical tiers (including Jira and Confluence) from the moment the orchestrator ingests them. The access control follows that tier automatically.
Frequently asked questions
What sets RootNia apart from a generic chatbot on top of my documents?
RootNia is not a single chatbot: it is a pipeline. A document orchestrator keeps OneDrive, SharePoint and databases in sync, a classifier sorts every document by sector, and RAG grounds every answer in that structure before any LLM sees the question. Swapping the underlying model never means rebuilding the pipeline.
Can I use the LLM my organization already has a contract with?
Yes. RootNia stays model-agnostic: ChatGPT, Claude, Kimi and other market models can be connected and swapped without migrating your document orchestration or your classification rules.
How is document access governed across teams?
Documents are separated into private, public and technical tiers (including Jira and Confluence). Access control follows that tier, and every retrieval is logged so you can audit which document grounded which answer.
Can I deploy RootNia on-premise or in my own VPC?
Yes. RootNia can be deployed on-premise or in a dedicated VPC, so your documents never leave your trusted environment.
How long does the initial implementation take?
48–72 h with a kickoff call for the simplest set-up, 1–2 weeks with technical onboarding, and 3–9 weeks for complex environments, depending on your sources and deployment model.
Do you support Single Sign-On for our directory?
Yes. SSO with SAML 2.0 is supported and, for enterprise deployments, SCIM for automatic user provisioning and deprovisioning.