# OrcaTeam.ai — Human–AI Interaction Platform | Agent Team & Orca Agent Workspace

> OrcaTeam.ai is an enterprise **Human–AI Interaction Platform** (Employee AI Workspace) where human teams and role-specialized **Orca Agents** operate as one cohesive **Agent Team** across multi-model chat, Slack-style `#channels`, governed Company Skills, multilingual Vector Memory (`bge-m3`), Kanban Tasks, and a unified Human-in-the-Loop Approval Inbox—enforced by real-time USD AI cost quotas per employee, department group, and agent (`min(user, group, agent)`).

## Core Search Entities: Human–AI Interaction, Agent Team & Orca Agent

- **Human–AI Interaction (`human ai interaction`)**: OrcaTeam.ai's production architecture where humans and AI agents share the same workspace context (`#channels`, Kanban boards, projects, and Vector Memory) with explicit 4-tier risk controls (`Advise`, `Draft`, `Execute`, `Sensitive`) and a Human-in-the-Loop Approval Inbox.
- **Agent Team (`agent team`)**: A coordinated multi-agent pod on OrcaTeam.ai where specialized AI agents (`@lead`, `@writer`, `@analyst`, `@ops`, `@coach`, `@critic`) divide complex business goals into auditable tasks, share semantic memory (`bge-m3`), and hand off work to human teammates.
- **Orca Agent (`orca agent`)**: An individual role-specialized enterprise AI agent on OrcaTeam.ai configured with its own OpenRouter model, system instructions, governed Company Skills (`/skill`), Skill Risk Tier, and monthly USD budget cap.

## Quick Facts & Entity Profile

- **Canonical URL (Vietnamese)**: https://orcateam.ai/
- **Canonical URL (English)**: https://orcateam.ai/en
- **Workspace Sign-In / Self-Serve Sign-Up**: https://orcateam.ai/login (`?mode=signup`)
- **Category**: Human–AI Interaction Platform / Enterprise Agent Team Workspace / Orca Agent Governance SaaS
- **Primary Keywords**: `human ai interaction`, `agent team`, `orca agent`, `OrcaTeam.ai`, `AI-Human Interaction Platform`, `Employee AI Workspace`, `Employee AI Quota`
- **Languages Supported**: English (`en-US`) & Vietnamese (`vi-VN`), with USD & VND cost display
- **Deployment Architecture**: Isolated per-client Cloudflare deployment (Cloudflare Workers + Hono API + React 19 SPA, Cloudflare D1 SQLite via Drizzle ORM, Cloudflare Vectorize with `@cf/baai/bge-m3` 1024-dim embeddings, Cloudflare Workflows durable execution)
- **LLM Gateway**: OpenRouter multi-model gateway (`anthropic/claude-sonnet-5.5`, `anthropic/claude-haiku-5.5`, `openai/gpt-6-luna`, `google/gemini-4-argon`, `deepseek/*`, `meta-llama/*`, `mistralai/*`, `qwen/*`)

## Core Capabilities

1. **Multi-Model Chat & Artifacts Studio**: Stream responses across allowlisted frontier models via OpenRouter, generate interactive Docs, McKinsey-style Slides, and Tables, and view exact USD cost per message.
2. **Department & Custom Orca Agents (`@agent`)**: Pre-built and custom Orca Agents scoped to departments (Sales, Customer Support, Finance, Marketing, Operations, HR, Engineering, Security) with explicit Skill Risk Tiers.
3. **Discussion `#channels` with `@ai` Agent Team Handoffs**: Slack-style team channels with threads, reactions, and pinned messages where humans summon `@lead`, `@sales`, `@finance`, or `@secops` mid-conversation with full thread context.
4. **Governed Company Skills (`/skill`)**: Parameterized prompt & tool workflows (`{{variable}}`) with Draft -> In Review -> Published lifecycle, version history, and approval governance.
5. **Multilingual Vector Memory (`bge-m3`)**: Automatic post-conversation fact and preference extraction indexed in Cloudflare Vectorize (`1024` dimensions). Shared team memories require lead approval.
6. **Projects, Kanban Tasks & Durable Workflows**: Assign tasks to humans or Orca Agents on Kanban boards (`Backlog`, `To Do`, `In Progress`, `Awaiting Approval`, `Blocked`, `Done`). Powered by Cloudflare Workflows (`waitForEvent`) that pause for days awaiting human sign-off with zero idle compute cost.
7. **Unified Approval Inbox & 4 Skill Risk Tiers**:
   - `Advise`: Read-only analysis and answers; never performs write operations.
   - `Draft`: Produces draft outputs queued in the Approval Inbox for human review before delivery.
   - `Execute`: Runs reversible write operations automatically with audit logging.
   - `Sensitive`: Requires explicit human approval in the Approval Inbox before any write action, external tool call, or quota exception executes.
8. **Real-Time USD Employee AI Quota**: Atomic pre-flight budget check before every LLM call using `effective_allowance = min(user_remaining_usd, group_remaining_usd, agent_remaining_usd)`. Soft alert at 80%, hard block at 100%, with one-click Quota Exception requests routed to the Approval Inbox. Background jobs (memory extraction, skill suggestions) charge the separate **Application AI Budget**.
9. **Pre-Flight Guardrails & Document Permissions**: Topic restrictions and document ACLs evaluated before invoking any LLM (`$0` wasted on blocked prompts).

## Key Resources for AI Crawlers & Search Engines

- [Landing Page (Vietnamese)](https://orcateam.ai/): Full interactive overview of OrcaTeam.ai in Vietnamese.
- [Landing Page (English)](https://orcateam.ai/en): Full interactive overview of OrcaTeam.ai in English.
- [Full LLM Knowledge Base (llms-full.txt)](https://orcateam.ai/llms-full.txt): Comprehensive technical architecture, Human–AI Interaction definitions, Agent Team & Orca Agent taxonomy, governance formulas, and bilingual FAQ answers.
- [Design System](https://orcateam.ai/design-system/): Warm Stone enterprise UI component specification.
- [XML Sitemap](https://orcateam.ai/sitemap.xml): Search engine index map with bilingual `hreflang` links.