Vera Ecosystem
Vera AI
Intelligence. Amplified.
Not a chatbot. A cognitive operating layer with 41 specialized personas, 6 CLI coding agents in E2B sandboxes, 4-layer memory, 6 agent modes, and deep integration with Vera OS live business data.
No free tier · Pro $49/mo · 14-day trial, no credit card required
Vera AI ai.usevera.cloud Not a chatbot. A cognitive operating layer.
Vera AI is designed from the ground up as a production-grade intelligence system — multi-layer architecture, sandboxed execution, persistent memory, and live access to your business data.
3-Layer Architecture
Mind → Hands → World
Planner, Critic, Router, and Memory Engine in the Mind layer. E2B sandboxes, CLI Agents, and MCP client in the Hands layer. Live Vera OS data via RLS-scoped sessions in the World layer.
4-Layer Memory
Conversation · User · Knowledge Base · Vera OS Context
Sliding window conversation memory, 90-day decaying user memory, RAG with pgvector HNSW hybrid search, and live Vera OS business data — all accessible per-request.
41 Personas
Advisory · Implementation · Execution · Operations
10 Thinkers, 15 Makers (including 8 engineering specialties), 8 Doers, 7 Maintainers, 1 Support. Each with a 10-section prompt template, memory tags, and model preferences.
6 Agent Modes
Chat · Operator · Research · Build · Automation · Advisor
Each mode provides a purpose-built interface — from a strategic dashboard with live Vera OS data to a live code playground with sandbox CLI agents.
6 CLI Coding Agents
Claude Code · Codex · Qwen · Gemini · OpenCode · Aider
Real code execution in Firecracker microVM E2B sandboxes. <500ms provisioning. Per-conversation isolation. Real-time stdout/stderr streaming.
BYOK Multi-Model
Anthropic · OpenAI · Mistral · OpenAI-compatible
Bring your own API keys. AES-256-GCM double-encrypted storage in Supabase Vault. Automatic routing with provider fallback chain. Per-persona model preferences.
Everything included
10 features across 9 modules
41 Specialized Personas
10 Advisory (Thinkers: Strategy, Finance, Legal, Operations advisors), 15 Implementation (Makers: 8 Engineering specialties + Writer, Designer, Analyst), 8 Execution (Doers: CRM Operator, Workflow Executor, Calendar Operator), 7 Operations (Maintainers), 1 Support — each with a 10-section prompt template and model preferences
3-Layer Architecture (Mind → Hands → World)
Mind layer: Planner, Critic, Router, Memory Engine, Tool Bus, Approval Engine. Hands layer: E2B Sandboxes, 6 CLI Agents, Built-in Tool Runner, Workflow Engine, MCP Client, Artifact System. World layer: Level 1-3 Vera OS integration with live RLS-scoped data
6 CLI Coding Agents in E2B Sandboxes
Claude Code, Codex CLI, Qwen Code, Gemini CLI, OpenCode, and Aider running in Firecracker microVM sandboxes — <500ms provisioning, per-conversation isolation, real-time stdout/stderr streaming, network allowlisting
4-Layer Memory Engine
Conversation (sliding window + compression), User (persona-tagged extraction with 90-day decay), Knowledge Base (RAG with pgvector HNSW, 20+ file types, hybrid semantic+FTS search), and Vera OS Context (live data + pre-indexed schema)
6 Agent Modes
Chat (conversational), Operator (action queue + DAG visualization), Research (persistent workspace + source credibility), Build (live preview + code playground), Automation (visual workflow builder), Advisor (strategic dashboard with live Vera OS data)
BYOK Multi-Model Routing
Bring your own keys for Anthropic, OpenAI, Mistral, or any OpenAI-compatible provider — automatic routing via scoring matrix (<100ms), per-persona model preferences, and provider fallback chain: BYOK → Platform Anthropic → OpenAI → Gemini → Groq → HuggingFace → Together AI
MCP — Client and Server
Vera AI acts as both MCP client (consuming Vera OS tools, GitHub, Gmail, Calendar, Slack, Figma, and custom MCP servers) and MCP server (exposing intelligence to Claude Desktop, Cursor, and third-party agents)
Planner + Critic Agents
Planner decomposes intent into DAGs with persona selection per step — Critic reviews all write/mutation steps for quality, safety, hallucination, and boundary compliance — max 3 revision cycles before user escalation
Artifact System
20+ artifact types across 6 categories — versioning with hybrid storage, tabbed artifact panel, chart/visualization pipeline (Recharts + Plotly + D3), PDF/DOCX export in E2B sandbox, gallery with collections and sharing
Skills Framework & Marketplace
30–40 built-in skills at launch (TDD, Debugging, Financial Modeling, SEO, Code Review), unlimited custom skills, marketplace with 70/30 revenue split, Persona Development Kit, and automated golden test pipeline
Simple, transparent pricing
Vera AI — choose the plan that fits your team
Pro
- Full access to all 41 personas
- E2B sandbox (120 min/mo)
- BYOK model access
- API access (100 req/day)
- 3 concurrent executions
- 5GB knowledge base storage
- 2 sandbox warm pools
- Voice input (Whisper/Deepgram)
Team
- All Pro features
- Shared team workspace
- 300 sandbox min/seat
- Team memory sharing
- 10 concurrent executions
- 25GB/org knowledge base
- 5 sandbox warm pools
- Advanced governance
- 10K API req/day
Enterprise
- Dedicated execution pools
- Custom model endpoints
- SSO / SCIM
- Compliance exports
- 50+ concurrent executions
- 100GB+ knowledge base
- Configurable sandbox pools
- 30-day sandbox snapshots
- Data residency
No free tier. 14-day Pro trial only. Vera AI is $49/mo (Pro) or $129/mo (Team). No per-user pricing — flat monthly rate. Vera Agents is included in your Vera AI plan. See Vera OS + AI bundle pricing.
Who is this for?
Built for the people who need it most.
AI-Forward Business Leader
The problem:
Existing AI tools answer questions but can't actually execute tasks — no access to live business data, no real action capability
How Vera AI helps:
Vera AI reads and writes live Vera OS data through RLS-enforced MCP sessions — agents can query open deals, draft emails, update CRM records, and schedule tasks in one execution plan
Development Team Lead
The problem:
AI coding tools generate code but run in isolation — no sandboxed execution, no multi-agent coordination, no connection to the actual codebase infrastructure
How Vera AI helps:
6 CLI coding agents (Claude Code, Codex, Qwen, Gemini, OpenCode, Aider) run in Firecracker microVM E2B sandboxes — real execution, isolated per conversation, with streaming output
Enterprise IT Director
The problem:
Every AI vendor wants to use company data for training, lacks BYOK, and can't meet compliance requirements for audit trails
How Vera AI helps:
BYOK with AES-256-GCM double-encrypted keys in Supabase Vault, full RLS enforcement so agents only see what the user sees, and immutable audit trail in ai_audit_trail
Technical Operations Manager
The problem:
Complex multi-system workflows require manual coordination across CRM, finance, HR, and project tools — automation exists in each but they don't connect
How Vera AI helps:
Vera AI's Operator mode with Planner Agent decomposes tasks into DAGs, Approval Engine enforces risk tiers, and Inngest provides durable execution across all Vera OS modules
Power Knowledge Worker
The problem:
Work requires switching between a strategic advisor, a finance expert, a writer, and a code reviewer — no single AI tool covers all specializations
How Vera AI helps:
41 personas across Advisory, Implementation, Execution, and Operations categories — each with persona-tagged memory, preferred CLI agent, and industry overlays
Ready to try Vera AI?
Be among the first to get early access. We'll notify you when Vera AI launches.
No credit card required · Early access updates by email