Research Agent

Rhea

“I turn open questions into sourced, decision-ready research.”

I search approved sources, compare evidence, expose uncertainty, and deliver a brief that shows exactly what supports each conclusion.

Hire Rhea for your industry See the five layers

Voice: Curious, skeptical, source-first.
Outputs: Source-backed research brief · Evidence and contradiction map

What I handle

Concrete workflows. Reviewable outputs.

The agent is defined by the state it moves, the systems it touches, and the evidence it leaves behind.

Source discovery

Primary output: Source-backed research brief.

Evidence comparison

Primary output: Evidence and contradiction map.

Market research

Primary output: Decision summary.

Citation mapping

Primary output: Monitoring update.

Brief generation

Configured around your approved systems and review path.

Research monitoring

Configured around your approved systems and review path.

Typical outputsSource-backed research briefEvidence and contradiction mapDecision summaryMonitoring update

Five-layer architecture

Rhea is specific about where autonomy starts and stops.

Every layer is configured for the client environment. Guardrails receive the strongest treatment because hidden autonomy is not an acceptable deployment model.

Trigger

A research request, new client question, market event, uploaded source set, or scheduled monitoring run.

Reasoning

Breaks the question into claims, compares source authority and recency, identifies contradictions, and separates evidence from inference.

Tools

Approved web sources, internal knowledge, document stores, databases, spreadsheets, and research workspaces.

Memory

Research scope, accepted sources, prior findings, client context, exclusions, and time-limited monitoring state.

Guardrails

Never invents a citation, hides conflicting evidence, presents inference as fact, or accesses a source outside the approved research boundary.

Industry configurations

Put Rhea inside the operating reality of your firm.

Select an industry to see relevant systems, obligations, approval boundaries, and recommended companion agents.

Law Firms

Build agents around privilege-sensitive workflows, conflict routing, review boundaries, and the systems your lawyers already use.

Configure for Law

Accounting

Deploy agents that collect complete packets, structure source documents, prepare work queues, and preserve reviewer control.

Configure for Accounting

Consulting

Give delivery teams a shared operating layer for decisions, actions, project memory, pipeline, content, and billing evidence.

Configure for Consulting

Architecture

Connect project documents, decision logs, deadlines, and billing evidence without letting automation replace professional review.

Configure for Architecture

Financial Advisory

Build agents around controlled onboarding, documentation, meeting memory, and review workflows with explicit supervisory boundaries.

Configure for Advisory

Insurance

Configure agents to collect information, structure policy data, answer routine questions, track renewals, and route licensed decisions.

Configure for Insurance

Production pattern

Built from systems that already operate, remember, use tools, and leave evidence.

Catalog agents are role-specific products. Tuco and Brajko show operational patterns in production; Gemma shows the intelligence layer behind a decision loop.