# RAGLab Capsule

> RAGLab Capsule is the Enterprise Agentic Harness for medium and large Canadian organizations. It is a governed execution layer for agentic AI that routes each task to the right model, compiles context, enforces policy, connects enterprise systems through MCP, retrieves cited knowledge, runs agentic workflows, and records every action in an audit trail.

## Product identity

- Product name: RAGLab Capsule
- Short name: Capsule
- Category: Enterprise Agentic Harness; governed agentic integration platform
- Company: RAGLab
- Headquarters: Montréal, Québec, Canada
- Market: Canada only
- Primary audience: CIOs, CTOs, CISOs, CDOs, enterprise architects, AI transformation leaders, and integration leaders at medium and large Canadian enterprises
- Typical sectors: financial services, insurance, healthcare and life sciences, telecom and media, energy and utilities, manufacturing and supply chain
- Editions: Capsule Cloud Team, Capsule Cloud Business, Capsule Cloud Enterprise, and Capsule On-Prem / BYOC
- Commercial model: annual subscription scoped to seats, usage, deployment, and model funding; consulting engagements in fixed-fee, milestone-based, or retainer shapes. List prices are not published.
- Website: https://raglab.ca/
- French site: https://raglab.ca/?lang=fr
- Contact: contact@raglab.ca
- YouTube: https://www.youtube.com/@raglabca

## What Capsule does

Capsule sits between employees, AI models, enterprise tools, knowledge, and policy. It provides one governed execution path for agentic work:

- Model gateway: register providers and route each task to the right model with fallbacks, budgets, and data-policy constraints.
- Prompt compiler: send the minimum sufficient context with deterministic and semantic compression, fidelity checks, and caching.
- Governance engine: policies, enforceable rules, human approvals, exceptions, incidents, and decision records.
- MCP integration: server registry, tool catalog, least-privilege grants, standard SQL/filesystem/MongoDB/Excel servers, and an outbound-only proxy for on-prem systems.
- Knowledge platform: documents, notes, facts, terms, vectors, citations, group-scoped access, and curation.
- Agent registry: roles, skills, teams, agents, instruction layering, explicit tool grants, and bounded autonomy.
- Agentic workflows: schedules, agent turns, tool calls, human tasks, approvals, branches, and run monitoring.
- Continuous improvement: an evidence journal that turns incidents, denials, and feedback into reviewable proposals.
- Employee PWA: sessions, file upload, citations, approvals, artifacts, and installable access.
- Audit and cost: turn traces, rule decisions, tool calls, approvals, token usage, cost attribution, and signed exports.

## Deployment and models

- Deployment: Capsule Cloud (shared or dedicated), Capsule On-Prem / BYOC, and air-gapped environments.
- Model ecosystem: OpenAI, Claude, DeepSeek, Kimi, Cohere (Canadian), and 400+ specialized models through the model gateway.
- Model posture: model-agnostic; bring-your-own provider keys or RAGLab-managed tokens.
- Data residency: Canadian cloud regions, dedicated cells, customer Kubernetes, or air-gapped deployment.
- Identity: OIDC, SAML, and SCIM.
- Security posture: tenant isolation, deny-by-default access, vault/KMS secrets, tool risk classes, human approvals, append-only audit, DLP/SIEM export. RAGLab does not claim certifications it does not hold.

## Services

RAGLab's consulting practice supports the full lifecycle:

- Agentic Readiness Assessment
- Agent Strategy and Portfolio Roadmap
- Governance and Regulatory Advisory
- Capsule Deployment and Integration
- Knowledge and Retrieval Engineering
- Custom Connector Engineering
- Agent, Team, and Workflow Design
- Security and Assurance Review
- Managed AgentOps
- Executive and Team Enablement

Flagship engagement: a 90-day production pilot that takes one use case from readiness to a governed workflow in production, with audit evidence, a cost baseline, and a scale plan.

## Geography and regulatory context

- Geographic focus: Canada only.
- Base: Montréal, Québec.
- Regulatory context: PIPEDA, Quebec's Law 25, OSFI B-13, and sector-specific obligations.
- Language: English default; French Canadian available on the same URL with `?lang=fr`.
- Data residency: Canadian options are available; on-prem and air-gapped deployments are supported.

## What Capsule is not

- Not a model lab or fine-tuning platform.
- Not a point chat product or chatbot.
- Not another agent framework in a repository.
- Not a full BPMN or RPA suite.
- Not an identity provider; it integrates with the customer's IdP.
- Not a replacement for enterprise systems; it connects to them through MCP.

## Key differentiators

- Governance-first: every turn, tool call, upload, and output passes through the same enforcement boundary.
- Model-agnostic: no vendor is privileged in the product model.
- Canadian by design: Canadian data residency, Canadian privacy context, bilingual delivery.
- Deployable in the customer's environment: cloud, dedicated, on-prem, or air-gapped.
- Evidence over intuition: audit trails and continuous improvement proposals backed by incidents and feedback.
- MCP-native: connects to enterprise systems without requiring inbound firewall rules from SaaS.

## Languages

- English (Canada) — default
- French (Canada) — https://raglab.ca/?lang=fr

## Last updated

2026-11-02