01 · Agents
Composable workers with declared role + tools.
Each agent has an explicit role, a permitted toolset (MCP servers, internal APIs, retrieval indexes), and a model assignment. Composition is YAML; deployment is a CLI command.
Product · AI & Automation · Production
Maria
Enterprise AI orchestration. Compose, govern, and observe AI agents across your organization — without sending prompts or responses outside your perimeter.
The problem
Generic AI chat does not survive contact with an enterprise.
The first AI pilot ships fast. It produces a chatbot, a summarizer, and a retrieval demo. Then governance asks: who ran what, against what data, and where did the prompts go? The vendor's audit log does not answer those questions, and the SaaS endpoint is in the wrong country. The pilot stops there.
Maria starts where that pilot ends. It is the orchestration plane that turns ad-hoc AI use into a governed capability — agents are composed once, authorized per role, and observed per tool call. Data residency is a deployment decision, not a vendor's policy.
How it works
Four primitives, composed.
02 · Tools & connectors
First-class agentic-protocol support, plus native connectors.
Tools are protocol-first. Native connectors for the enterprise platforms, productivity suites, geospatial systems, and SQL/NoSQL stores most organizations run. Custom connectors are a few hundred lines, not a few weeks.
03 · Governance
Per-agent role authorization. Per-tool-call audit.
OIDC / SAML role mapping. Every tool call is logged with prompt context, parameters, and result. Logs are append-only and exportable for review.
04 · Observability
Latency, cost, and quality per workflow.
Standard metrics, distributed traces, and per-agent quality dashboards. Cost attribution by team, by workflow, by model.
Sovereign by default
Maria runs in your tenant — not ours.
Maria deploys as a containerized stack inside the customer's sovereign cloud tenant or on-premises. The control plane and the data plane both live in the customer's perimeter. There is no Maria SaaS — there is no shared cloud where customer prompts could end up.
Model inference uses whichever provider fits the use case: a Canadian-region frontier model, or an open model on the customer's own GPU. Maria knows which model is in scope per agent and routes accordingly.
Patterns that use this product
Worked examples from real engagements.
Government · Pattern
How we shipped AI workflow automation inside a Canadian municipality's tenant
Natural-language workflow building (Kacha) deployed inside the customer's tenant — no data leaves the perimeter, no license per workflow, no foreign cloud. From scoping to first workflow in production in under ten weeks.
8–12 weeks from scope to production
Telecom · Pattern
Agentic AI on a thirty-year-old OSS, without ripping it out
How we wired an agent runtime into a telecom operator's legacy customer system using a custom protocol connector and reusable skill files — turning a read-only audit nightmare into a queryable, governable AI workspace for operations teams.
8–12 weeks for a production read-access pilot
Get started
We deploy. We govern. You own the agents.
Tell us what you'd put an agent to work on. We'll come back with a 30-day deployment plan and a governance model your CISO will sign off on.
Talk to an engineer