One workspace - every model, tool, and channel

Build Enterprise AI Agents That Ship to Production

OneAI Agent Craft is the agentic OS for the enterprise. Configure an agent's model, knowledge, tools, and guardrails once-then let smart routing, durable memory, and full governance carry it from a prompt to a monitored, versioned deployment.

No credit card required · SSO & SCIM ready · Deploy on your cloud or ours

oneai.rhatechnologies.com/agents/finance-analyst/build

Configure Agent

Routing Engine Status

Auto - Model Planner picks optimal model per turn

Simple lookupHaiku · Lite

Reasoning & analysisClaude 3.5 Sonnet / Opus

Grounded web searchsearch_model

Active Agent Capabilities

Agentic MemoryCitationsWeb searchArtifactsData analysisEmail + Calendar

Live Test Previewv5 Candidate

Summarise Q3 spend vs budget and flag overruns.

Q3 spend was $4.18M - 6.2% over budget. Marketing (+18%) and Cloud Infra (+11%) drove the primary overrun.

3 sources cited · Finance KB v2

Runs on the LLMs you already trustAnthropic ClaudeAzure OpenAIAmazon BedrockGoogle Vertex AIOllama (Self-Hosted)

Production-Grade Workflow

From Prompt to Production in 5 Governed Steps

No glue code, no stitched-together vendors. Every stage of the agent lifecycle lives in one place - with governance and cost controls on by default.

Define the agent

Name, model, system prompt, knowledge bases, and tools. Assign seat tiers to inherit guardrails.

Turn on capabilities

Toggle memory, citations, web search, artifacts, voice, data analysis, email & calendar with 1 click.

Test & evaluate

Run routing datasets and quality evals against a version before promoting. Compare side by side.

Deploy across channels

Publish to Teams, Slack, email, calendar, REST SDK, MCP/A2A, or desktop app from one build.

Monitor & scale

Watch latency, tokens, and cost per chat in real time. Roll back instantly and iterate securely.

Unified Agent Control Plane

One Control Plane for Every Model, Tool & Surface

A single request flows through routing, memory, knowledge, tools, and guardrails - governed and cost-attributed end to end.

Highlight:

Auto Planner

Model Routing & Execution

Model Planner evaluates prompt intent and routes dynamically - pro model for reasoning, lite model for lookups, or search model for live web.

Latency: ~320ms

Supported: Claude, Azure OpenAI, Bedrock, Vertex

Dynamic Temperature

Unified Configuration Across Models, Knowledge & Actions

Orchestration, knowledge, and tools aren't add-ons - they're wired into the runtime, governed by the same policies, and attributed in the same cost ledger.

Orchestration & routing

Multi-agent handoffs, deterministic precondition triggers, and a Model Planner that picks the right model per turn - pro for reasoning, lite for lookups, search model for live web.

Knowledge & grounding

Retrieval over approved knowledge bases with inline citations, tabular datasets, and cold-doc handling. Answers are grounded strictly in approved corporate context.

Tools, skills & MCP

Connect email, calendar, Teams, Jira, and HRMS as first-class capabilities, or bring any Model Context Protocol (MCP) connector with per-tenant permission scoping.

Instant Capability Toggles with Zero Redeployment

Each switch changes real runtime behavior - no redeploy needed, governed by the agent's assigned seat tier.

Agentic Memory

Remembers preferences, people, and prior work across sessions.

Inline Citations

Traceable source links on every grounded claim or doc query.

Live Web Search

A dedicated search model grounds answers on live results.

Artifacts Engine

Documents, slide decks, and charts rendered as editable cards.

Data Analysis & AutoBI

Upload CSV/XLSX for automated datasets and visual dashboards.

Voice Intelligence

Real-time speech transcription and natural voice response.

Email & Calendar

Triage inbox, draft replies, and schedule with confirm cards.

Tasks & Follow-ups

Native task sync with Microsoft To-Do and Google Tasks parity.

Enterprise Guardrails

Prompt-injection defense, CBAC scoping, and confirm-gated actions.

Agents That Remember, and Get More Reliable Over Time

OneAI extracts durable memory automatically as people work - preferences weighted twice, people and projects linked in a Work Graph, feedback closing the loop. Anaphora like "reply to that one" resolves because the agent keeps a referent registry across turns.

Persistent memory across sessions, users, and channels
A connected Work Graph of people, threads, and commitments
Governance built in - isolation, policy, and audit on every memory

Explore Agentic Memory

What OneAI Remembered

Prefers concise, bulleted summaries

User preference · Weighted 2×

"The board deck" → Q3 Board Review workstream

Connected memory · Work Graph

Send test emails to team@rhaone.ai

Standing rule · Applied automatically

Deploy the Same Agent Everywhere

One version, many surfaces. Publish once and OneAI keeps every channel in sync.

REST API & SDK

# OneAI SDK - Run your agent programmatically
curl -X POST https://oneai.rhatechnologies.com/v1/agents/finance-analyst/run \
  -H "Authorization: Bearer $ONEAI_KEY" \
  -H "Content-Type: application/json" \
  -d '{"input": "Q3 spend vs budget analysis"}'

MCP & Agent-to-Agent (A2A)

# Expose your agent as an MCP tool
{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "finance-analyst",
    "arguments": { "quarter": "Q3" }
  }
}

Microsoft Teams

Agent DMs & channels, send-as-user, mobile-ready cards.

Email & Calendar

Triage, drafts, and scheduling with a 'Sent via OneAI' footer.

Embed Widget

Drop a single script tag into any internal portal or web app.

Desktop App

Native macOS & Windows desktop builds powered by Tauri.

Build Enterprise AI Agents Today

Configure models, durable memory, and guardrails in minutes. Deploy securely across your organization.