Product Roadmap
What's Built
Shipped and running in the current release
Comprehensive Data Model
17 models and 18 enums on Prisma + PostgreSQL, covering organizations, projects, tactics, tasks, metrics, agents, and documents.
Eight Service Layers
Organization, Project, Tactic, Task, Metric, MetricsConfig, Agent, and Document services with 29 REST API routes.
Garden Visualization UI
Living SVG organisms that represent projects — branches, leaves, and health bars that shift and grow as campaigns progress.
Agentic Task System
Eight specialized gnomes (Social Media, Community, Store Presence, Content Marketing, Designer, Research, Videographer, General) with a full plan → approve → execute lifecycle powered by the Anthropic API.
Project Knowledge Base
Brand voices, product briefs, audience profiles, and content guidelines injected directly into agent context so their work stays aligned.
MCP Server Integration
Model Context Protocol server for Claude Desktop and Claude Code — manage your garden conversationally without switching tabs.
Credential Vault
AES-256-GCM encrypted storage for API keys and platform credentials, keeping agent integrations secure.
MVP Frontend
Project detail views, tactic and task management, agent assignment, and real-time execution monitoring.
Slack Integration
Slack platform tools for posting, reading threads, and searching channels — plus a thread-based approval flow where reviewers type approve, reject, or revise to drive work-product review.
Nano Banana Image Generation
Gnomes can produce visual work products via Google's Nano Banana image model, flowing generated media into the project library and onto downstream deliveries.
Platform Connectors
Production-grade platform tools for Twitter, Steam, Discord, Google Analytics, Instagram, App Store, Slack, and web search — wired into gnomes with credential vault and approval flows.
Video Production
Videographer gnome produces explainer videos via the Lenzon rendering pipeline, with async fire-and-return and webhook-driven completion.
Zernio Multi-Channel Delivery
Org-keyed Zernio integration with per-project profiles delivers finished work products to downstream channels — YouTube publishing live.
Scheduled Task Execution
Cron-driven tactic runs, automated metric pulls, scheduled gnome reviews, and background sweepers for long-running sessions and approvals.
Near-Term
Actively in development or next in the queue
Analyst Gnomecurrent focus
A new specialized gnome that interprets each project's goal, sets up the right measurement (GA events, manual logs, alternate sources), proposes tactics tuned to the current trend, and renders a project-specific chart that narrates what's actually moving.
Task Recommendations
Context-aware task suggestions when you create a tactic — pre-tailored to your brand, audience, and platform.
Tactic Templates
Pre-built playbooks like “Game Launch” or “Instagram Growth” with suggested tactics and task templates to reduce cold-start friction.
Execution Visualization
Real-time step progress, streaming output, timeline views, and clearer visual distinction between read-only and side-effect steps.
Medium-Term
Planned once near-term foundations are solid
Multi-Org Dashboard
Org-scoped views, org switcher, member management, and shared credential administration.
Notifications & Alerts
Agent completions, metric threshold crossings, and health drops surfaced in-app and via Slack, Discord, or email.
Cost & Usage Tracking
Per-project, per-agent, per-org dashboards for token spend and execution duration — budget visibility for multi-tenant environments.
Long-Term Vision
Where the garden grows next
Agent Marketplace
A community marketplace for specialized agents — indie game marketing, SaaS email sequences, e-commerce retention.
Cross-Project Intelligence
Pattern recognition across projects: “This tactic worked for Project A — consider it for Project B.”
External Integrations Ecosystem
Notion sync, Google Docs, Zapier webhooks, Slack bots — marymary as the natural hub for marketing operations.
Self-Improving Agents
Agents that adapt their approach over time using execution history and outcome data, growing stronger with every run.