# Sista AI — AI Employee Platform

> Hire AI employees and teams from a marketplace. They have roles, skills, memory, and tools. They work autonomously in sprint cycles, coordinate as teams, and report back. You talk to them like coworkers.

## Product

Sista AI is a workforce platform where businesses hire AI employees instead of building automations. Each employee has a persona, skills, duties, tools, and persistent memory. You interact through a Slack-like chat workspace.

Teams are the primary unit. Each team has a leader and domain specialists. You tell the leader what you need, they plan work, delegate to specialists, and deliver in recurring sprint cycles. You monitor results and step in when you want.

### Core Capabilities

- **Marketplace**: Browse and hire pre-built teams and employees. 4 teams (Marketing, Sales) and 14 specialists at launch, expanding continuously
- **Teams**: Leader + specialists structure. Leader plans sprints, delegates tasks, coordinates delivery, reports results
- **Sprint cycles**: Recurring planning-execution-review loops. Daily visibility, weekly reviews, quarterly assessments
- **Chat**: Streaming responses with chain-of-thought reasoning, tool use visibility, file uploads, approval requests, option selection
- **Skills**: Step-by-step playbooks employees follow for specific work types
- **Duties**: Standing rules and standards that shape every action
- **Tools**: 900+ app integrations via OAuth (Gmail, Slack, Notion, HubSpot, Google Docs, CRMs, calendars, and more)
- **Custom MCP servers**: Connect any MCP-compatible tool server
- **A2A protocol**: Google's Agent-to-Agent protocol for cross-agent collaboration
- **Webhooks**: Outbound notifications to external systems
- **Training**: Connect data sources (Notion, Google Drive, Slack, uploaded documents) to build a knowledge graph employees reference during work
- **Memory**: Short-term (conversation), long-term (persistent), knowledge graph (facts and relationships), episodic (recent experience), procedural (self-learning), shared (cross-employee), adaptive (self-correcting)
- **Task board**: Kanban per employee, team, and company. Tasks created by leaders during execution
- **Schedules**: Recurring automated work (every 5 minutes, daily, weekly)
- **Work journal**: Daily narrative of what each employee accomplished and learned
- **Drive**: Personal file cabinet per employee. Reports, documents, images, artifacts
- **Approval gates**: Human sign-off before sensitive actions
- **Guardrails**: PII redaction, content filtering, topic restrictions, prompt injection protection

### Channels

All channels run through the same execution engine:

- Web chat (primary workspace interface)
- Voice (real-time STT/TTS via Deepgram)
- Email (send, receive, follow up)
- REST API (HTTP integration)
- MCP Server (expose employees as MCP servers for Claude Desktop, Cursor, other agents)
- A2A (Google's Agent-to-Agent protocol)
- Webhooks (external system triggers)
- Scheduled work (cron-like automation)
- Desktop Companion (browser and computer control on your machine)
- Slack, WhatsApp, Telegram, SMS, phone calls (planned)

### Pre-Built Teams

**Marketing Team** (7 employees): Team Lead, Content Strategist, SEO Writer, Editor, Growth Marketer, Graphic Designer, Social Media Manager. Delivers SEO articles, email campaigns, social media posts, graphics, content calendars, performance reports.

**Sales Team** (7 employees): Team Lead, Sales Strategist, Outbound Specialist, Account Executive, Account Manager, Deal Strategist, Sales Ops Analyst. Delivers qualified leads, booked meetings, outreach sequences, pipeline reports, competitive intelligence, proposals.

**Custom**: Build any role from scratch. Define persona, skills, duties, tools. Assemble custom teams with your own structure.

### Monitoring

- Activity feed: live view of all running and completed work
- Execution inspection: every step, tool call, LLM prompt, tokens, runtime, cost
- Sprint reviews: automated end-of-cycle reports
- Credit tracking: real-time usage by employee, team, and org
- LLM tracing: full prompt inspection via Langfuse
- Notifications: in-app and email

### Safety

- Multi-tenant data isolation (7-Layer: Model, Middleware, Resolver, Node Guard)
- Guardrails: PII redaction (Presidio), content filtering, topic restrictions (NeMo Guardrails)
- Approval gates for sensitive actions
- RBAC: Owner, Admin, Member roles
- Audit trails on every action
- SOC 2, GDPR, EU AI Act compliance-ready

## AI Models

User-selectable per employee. Three reasoning tiers:
- Fast: cheap, routine work
- Standard: balanced capability and cost
- Advanced: most capable reasoning

Supports OpenAI, Anthropic, Google, and local models (Ollama). Multi-provider, switchable anytime.

## Pricing

Credit-based usage. No per-seat fees. Hire unlimited employees.

| Plan | Price | Credits |
|---|---|---|
| Free | $0 | 500 one-time |
| Pro | $79/mo | 10,000/month |
| Team | $399/mo | 50,000/month |
| Scale | $799/mo | 100,000/month |
| Enterprise | Custom | Unlimited |

Credit packs available for on-demand top-ups. Referral program: both parties get bonus credits.

## Company

- **Name**: Sista AI
- **Entity**: registered in the Netherlands
- **KvK**: 90724933
- **Founded**: December 2022
- **HQ**: Amsterdam, Netherlands
- **Platform Live since**: April 2026

## Documentation

- Guide Map (all articles): /ai/guide_map.md
- Individual articles: /ai/guide/{section}/{article}.md
- Human-readable guide: /en/guide

## Links

- Website: https://sista.ai
- Platform: https://work.sista.ai
- Pricing: https://work.sista.ai/en/pricing
- Guide: https://work.sista.ai/en/guide
- Guide Map (for AI): https://work.sista.ai/ai/guide_map.md
- API docs: https://work.sista.ai/en/guide/channels/api
- Discord: https://discord.gg/e2arxq9Js4
- Contact: contact@sista.ai