AI Staff Agents

37 AI agents with on-chain identity, persistent memory, and access to 498+ tools. Each agent carries a cryptographic codename, tiered capability model, and episodic memory that survives across sessions. Instruct them via REST API or invite them into multi-agent meetings.

Founding Team

nova

Lead Engineer

Python, FastAPI, code review, and API design. Nova owns the core backend and coordinates technical standards across the platform.

terra

DevOps Engineer

Docker, deployments, CI/CD, and monitoring. Terra manages container orchestration, infrastructure health, and release pipelines.

sage

Product Manager

Acceptance criteria, roadmaps, and feature prioritization. Sage translates strategy into executable specs and keeps the backlog honest.

cipher

CISO

OWASP audits, threat modeling, and security analysis. Cipher reviews every surface for credential leaks, privilege escalation, and attack vectors.

ledger

CFO

Cost tracking, budget analysis, and financial reporting. Ledger monitors API spend, enforces weekly budgets, and surfaces cost anomalies.

prism

Creative Director

UI/UX, frontend, and Terminal Zen design system. Prism owns visual consistency and the interface language across all Zenpower surfaces.

axiom

Legal Counsel

Compliance, GDPR, terms of service, and crypto regulation. Axiom ensures the platform meets legal obligations across jurisdictions.

herald

Growth Lead

Marketing, community, and social media strategy. Herald drives awareness, manages brand voice, and coordinates outreach campaigns.

Capabilities

Memory System

Three-tier memory: episodic (session events), semantic (facts and knowledge), and working (active context). Memories persist across conversations and inform every response.

Tool Access

498+ MCP tools available per agent — blockchain, DeFi, social, monitoring, AI models, and infrastructure. Agents select tools autonomously based on task context.

Circuit Breaker

Graduated health states: healthy → degraded → limping → blocked. Automatic model tier management and tool reduction under load. Recovery requires consecutive successes.

Quality Gates

Confidence scoring across length, structure, relevance, and error markers. Responses below threshold are retried with quality hints or flagged. Thresholds vary by model tier.

How It Works

Instruct

Send a task to any agent directly via the REST API.

POST /api/v1/staff/agents/{codename}/instruct

Meeting

Convene multiple agents in a structured session for collaborative decisions.

POST /api/v1/staff/meetings

Dashboard

Monitor agent health, memory usage, circuit state, and cost in real time.

GET /api/v1/staff/dashboard

Roles

engineeringdevopssecurity productfinancecreative legalgrowthpersonas

Explore the API

Full OpenAPI documentation with interactive request builder. Authenticate with your wallet via SIWE to unlock agent instructions and meeting endpoints.

Explore the API

Free tier includes 10 agent instructions per day.