The brain of your organization
AgentStaQ gives operators a central AI brain that routes work, streams reasoning, calls tools, and turns approved runs into reusable AI employees.
Back to products
AgentStaQ AI
The secure AI workforce platform for operational teams.
AgentStaQ is the brain of an organization: a living intelligence layer that connects business systems, orchestrates agents, captures human feedback, and turns proven work into deployable AI employees.
- 10
- Core product modules
- 3
- Connector modes
- 4
- Model routing tiers
- 24/7
- AI workforce runtime
Product vision
AI employees that learn from the way the company actually works.
AgentStaQ is designed to move AI from one-off chat into repeatable operations. The product treats every successful workflow as a future employee: reviewed, versioned, connected, trained, scheduled, and measured.
The long-term vision is an organization-level brain where documents, connector activity, staff corrections, approved outputs, and run history become institutional memory. New agents start with context, existing agents improve through review, and recurring work gets cheaper as the system learns.
Overview
One platform for agent creation, execution, review, and deployment.
Connects to any system
The connector layer supports MCP tools, direct REST and OAuth APIs, browser automation with MFA, and custom adapters for proprietary systems.
Human-in-the-loop by design
Outputs land in Review by default. Staff approve, correct, and promote high-quality runs into named employees with schedules, triggers, and versioned profiles.
Built for regulated work
The platform is designed around role-based access, encrypted credentials, self-hosted model options, audit logs, and blockchain-anchored run history.
Features
Product modules built around real operating workflows.
Modules map to the AgentStaQ walkthrough: Home, Brain, Connectors, History, Review, Employees, Notifications, Security, and the Company Brain.
Home dashboard
Live metrics for total runs, pending reviews, approved work, background jobs, agent performance, cost per run, and organization brain growth.
Brain runtime
A master agent thread with Instant, Agent, Auto, and Train modes. Auto mode chooses the fastest and cheapest path for each task.
Connector registry
MCP, API, browser agent, and custom connector modes let agents work across EHRs, payer portals, finance tools, CRMs, file systems, and internal apps.
History and audit trail
Every run records status, timing, model, confidence, connector, token usage, user, and a SHA-256 audit anchor for tamper-evident verification.
Review queue
Human reviewers see the result, reasoning, and connector trail before output takes effect. Corrections become labelled training pairs.
AI employees
Approved workflows become named employees with three clear actions: Utilize for one-off work, Deploy for recurring jobs, and Train for updates.
Notifications
Bell alerts, Microsoft Teams routing, and webhook-style notifications keep the right team informed when runs finish or need review.
Company brain
Org-level RAG indexes SOPs, payer rules, approved outputs, corrections, and institutional memory so new agents inherit context from day one.
Security and compliance
Designed for healthcare, finance, and other regulated operations.
AgentStaQ is built with regulated workflows in mind: protected data, role-gated actions, encrypted connectors, traceable runs, and deployment options that keep sensitive inference close to the client environment.
Role-based permissions by user, business unit, title, and agent visibility.
AES-256 encrypted connector credentials, OAuth tokens, API keys, passwords, and TOTP seeds.
TLS 1.3 for connector calls, API endpoints, and browser agent sessions.
Self-hosted Tier 1 models for sensitive workloads where PHI should stay inside the client perimeter.
Append-only operational logs with blockchain-anchored hashes for run-level verification.
Deployment options for healthcare, finance, IT operations, white-label partners, and enterprise custom editions.
Tech stack
Production-grade infrastructure from day one.
The stack emphasizes security, cost control, open-source leverage, progressive learning, and no black-box dependency for day-to-day work.
Frontend
- React + Vite SPA
- Token-by-token streaming responses
- Background process indicators
- Real-time run status without refresh
Backend
- Express.js microservices
- FastAPI Python services
- LangGraph workflows
- LiteLLM model router and fallback proxy
Data layer
- MongoDB for users, logs, history, connectors, and employee profiles
- Supabase pgvector for brain memory, embeddings, RL corpus, and org RAG
- Redis semantic cache and run-result cache
- Polygon or Hyperledger audit anchors
AI and models
- EC2 + SageMaker for Qwen2.5, Phi-4, and Mistral local model tiers
- DeepSeek and Groq routing for low-cost hosted inference
- Claude Haiku, Claude Sonnet, GPT-4o-mini, and GPT-4o escalation tiers
- Monthly fine-tuning loop with HITL training pairs
Automation
- MCP tool calls
- REST and OAuth APIs
- Playwright or Stagehand browser agents
- Tesseract OCR for document ingestion
Security
- RBAC and least-privilege controls
- Encrypted connector registry
- Private VPC deployment options
- Audit export with verification links
Product information
Built as a platform, not a single-purpose bot.
Primary users
Operations, billing, clinical, IT, finance, compliance, and partner teams
Core workflow
Create a run in Brain, review output, save it as an employee, then deploy it on a schedule or trigger
Deployment model
Hosted, private infrastructure, white-label, and custom enterprise editions
Differentiator
The company brain compounds knowledge across agents, employees, corrections, and approved work