TrueNext Platform

The runtime that powers
every TrueNext agent.

MARO is TrueNext's proprietary multi-agent runtime. It handles orchestration, constitutional security, session management, and delivery — so every agent we ship is production-ready on day one.

A platform, not a product.

MARO — Multi Agent Reasoning Orchestrator — is the intelligence layer that every TrueNext agent runs on. It's not sold separately; it's how we guarantee quality, security, and consistency across every agent we ship.

User
CLI prompt · Web UI · API call
constitutional enforcement · integrity filtering
MARO Runtime
Controller Agent · Security · Session
specialist routing · parallel execution
Specialist A
Domain expert
Specialist B
Domain expert
Specialist N
Domain expert

Agents at work.

support-agent — powered by MARO
# CLI mode — direct prompt
$ support-agent "My package hasn't arrived, order #44821"
I've located order #44821. It was delayed at the regional hub
but is now out for delivery — expected today before 8 PM.
# persistent conversation context
$ support-agent "Can you escalate this to your team?" -c
Escalated. You'll receive a confirmation within 15 minutes.
# web UI mode
$ support-agent serve --port 443
support-agent → https://0.0.0.0:443

A constitution every agent
is bound to honor.

MARO's security model is built into the agent's identity — not added on top. Two independent layers, always active, never bypassable.

Layer 1 · Intent

Constitutional mandate

Every MARO agent operates under a behavioral constitution — a set of domain-specific policies that define what it will and will not engage with. Before any request reaches the reasoning layer, MARO evaluates it simultaneously against every constitutional policy. The agent never sees a request that conflicts with its mandate.

  • Each policy is a distinct, independently enforced boundary
  • All policies are evaluated simultaneously — no sequential bottleneck
  • Every policy must be satisfied for the request to proceed
  • A single violation refuses the request — before any reasoning begins
  • The constitution is composable — new boundaries can be added at any time
✓ Constitution honored → request proceeds
✗ Constitution violated → request refused
Layer 2 · Response

Integrity filter

Every response an agent produces passes through an integrity filter before it reaches the user. MARO ensures the agent's output stays within the boundaries of what it should share — sensitive context, confidential information, and system internals are detected and suppressed automatically.

  • Applied to every response, on every turn, without exception
  • Sensitive content is detected before it leaves the reasoning layer
  • Suppressed responses are replaced with a clear, safe message
  • The conversation continues normally — no disruption to the session
  • Both primary and secondary response streams are filtered independently
✓ Integrity confirmed → response delivered
✗ Violation detected → response suppressed
A living constitution. An agent's behavioral mandate is not fixed at birth — it grows with your use case. New policies can be introduced at any time, each enforced independently and simultaneously alongside existing ones.

Everything the runtime handles.

🧠

Controller + specialist model

A root controller agent intelligently routes each request to the right domain specialist. Add specialists to expand your agent's expertise across any domain — with no changes to the core runtime.

🌐

Built-in web UI

Every MARO agent includes a full chat interface on demand — no frontend infrastructure required, no separate deployment, no additional configuration.

💬

Session persistence

Full multi-turn conversation history with automatic session management. Persistent or stateless — configured per invocation, not per deployment.

🔒

Zero surface area

Agent logic and instructions are sealed within the MARO runtime. Nothing is accessible or readable by end users — no exposed configuration, no visible internals.

Simultaneous policy enforcement

All constitutional policies are evaluated in parallel — security adds no cumulative latency. More policies means more coverage, not slower responses.

🔗

Provider agnostic

Supports Anthropic and any OpenAI-compatible API — Ollama, local models, hosted providers. Constitutional enforcement and agent reasoning can use different providers independently.

Agents powered by MARO.
Built for your use case.

We design, build, and deliver purpose-built agents — tell us what your team needs.