Product Roadmap
We are moving quickly to build the world’s most robust open-source local AI runtime and managed tool ecosystem. Our goal is to empower developers to create autonomous, secure, and production-viable software agents.
Next 3 Immediate Moves 🚀
To catalyze traction and solidify our architectural authority, we are executing these three strategic actions immediately:
- Open-source the OSS Core NOW: Ship the
@privane/engine,privane-cli, standardized schemas, local sandboxed execution runtime, and SDKs. Momentum and developer trust are prioritized over absolute feature completeness. - Record & Package the Magic Demo: Formally capture the PR standup workflow (local Gemma compute -> secure GitHub scan -> hosted browser extraction with accessibility DOM pruning -> synthesised standup stand-in -> Slack dispatch). Package this directly as a clean example repository for developers to fork.
- Initiate Strategic Infrastructure Outreach: Reach out to critical infrastructure partners whose categories are showcased by our decoupled architecture, specifically Browserbase and PinchTab for browser virtualization, and Together AI, Groq, and Qdrant for localized edge/inference scale.
Product Milestones & Phases
Phase 1: OSS Core & CLI (Active Release)
Establishing the local-first execution runtime.
- WebGPU Accelerated Runtime: Zero-dependency fast local inference running directly in browser sandbox memory.
- Quantized GGUF inference: Direct support for custom model weights via system command runtimes.
- OpenAI-Compatible server daemon (
privane serve): Full SSE completions streams support. - Local Sandboxed Execution: Safe filesystem folders and SQLite query keywords filtering.
Phase 2: Privane Cloud & Managed Infrastructure (In Progress)
Scaling execution capabilities for enterprise agent workloads.
- Hosted Virtual Browser Nodes: Cloud browser session virtualization integrated directly with Browserbase & PinchTab.
- Transient OAuth Gateway: Safe, stateless connection hub managing SaaS authentication storage, refresh sessions, and permission scopes for Slack, GitHub, Gmail, and Notion.
- Real-time DOM Pruner: Extraction engine that prunes raw layout HTML into structured accessibility schemas cloud-side, cutting token transmission costs by over 95%.
- Metering & Operational Tracing: Tracing pipeline, usage analytics, billing triggers, and rate-limiting enforcement.
Phase 3: The Privane Hub & Multi-Agent Network (Planned)
Building the decentralized hybrid marketplace for AI execution.
- Registry Marketplace: The “npm for AI tools”, cataloging standard schema schemas, custom MCP configurations, and workflow scripts.
- Workspaces & Collaboration Hubs: Multi-tenant credential scopes, shared team API keys, and group billing metrics.
- Enterprise Self-Hosted Gateway: VPC deployment packages enabling completely private enterprise tool routing.
- Multi-Agent Orchestration Protocol: Direct orchestration frameworks to coordinate tasks across several local models and tools.