A full product team, as software. Every role runs a professional flow.
One orchestrator routes work across five role-flows. Each flow moves through the same gated phases. A human approves every transition. That approval log IS the audit trail.
| Role | What its flow runs | Agents in the flow |
|---|---|---|
| Product Manager | Market research → PRD → validation → go-to-market. | 31 agents |
| Tech Lead | Architecture → system design → sprint breakdown → release. | 26 agents |
| Developer | Technical spikes → design → estimation → implementation. | 21 agents |
| QA Engineer | Test strategy → coverage plan → automation → sign-off. | 20 agents |
| Data Scientist | EDA → feasibility → experiment plan → ML pipeline. | 26 agents |
Seamless knowledge transfer from any role to any role — every flow shares the same context.
Built and deployed today — orchestrator + all five flows live across ~8 services, used daily by the founder.
Current philosophy
Proactive approach with reactive action can produce deterministic quality.
The Spine — every flow runs through the same gates
- WHY before work — A structured WHY interview runs before any agent touches the task.
- Veto mid-flight — Reflection checkpoints pause the pipeline and wait for a human verdict.
- Phase approval — Work that belongs to a later phase is refused, not improvised.
Every other harness makes an individual faster. Current governs an organization.
| The Current Platform | Hermes | OpenClaw | CrewAI | DeerFlow 2.0 | MetaGPT / ChatDev | |
|---|---|---|---|---|---|---|
| Built for | Enterprise product teams | Personal self-improving agent | Personal multi-channel assistant | Devs building automations | Autonomous research runs | Research demos of AI software teams |
| Methodology | 5 role-flows × gated phases | None - one loop | None - opt-in pipelines | None - role labels only | None - lead-agent loop | Fixed role-play script, no gates |
| Human-in-the-loop | WHY intake, veto, phase approval | None | Opt-in approval gates | Opt-in task hooks + webhooks | Mid-run interrupts only | None - fully autonomous |
| Traceability | Vault, quality scores, Jira sync | Session search | Markdown files | AMP Traces dashboard | Outputs only | Generated artifacts only |
| Delivery | Hosted, managed SaaS | Self-host | Self-host | Framework + hosted AMP | Self-host | Self-host (open source) |
| Who operates it | PM, lead, QA, DS, DEV - the whole team | The individual | The individual | Engineers + no-code Studio | Researchers & engineers | Researchers & hobbyists |
| What happens to a human "no"? | Captured as a Delta. Compounds. | Evaporates | Evaporates | Evaporates | Evaporates | Evaporates |
Every competitor cell checked against their docs, GitHub, and product pages (July 2026, cited in Business Plan).
The moat is method + governance, not code. Agents are swappable. Adoption is not. You cannot clone a past — every veto and approval compounds into an asset no one can clone in a weekend.
When Claude takes the tactical work, Current runs the strategic.
Claude became how you write code with AI. Current becomes how an organization runs its entire product process with AI.
Investors
Current Labs is raising a seed round. Full market, model, and financials in the deck.
Ilan Dahan, Founder — ilan.dahn@currentlabs.dev
More from Current Labs
- Current Plugin for Figma — audits, scores, and exports design system components as production-ready code.
- AI.D Methodology
Blog
-
Getting Started with the Current Plugin — A Step-by-Step Tutorial
A hands-on walkthrough of installing and using the Current Plugin Figma plugin to extract design tokens, classify components, and generate production-ready code.
-
Extracting Design Tokens from Figma — A Complete Guide
Learn what design tokens are, why they matter for design systems, and how the Current Plugin automates token extraction from Figma into your codebase.
-
The Best Design-to-Code Tools in 2026 — A Landscape Overview
A comprehensive comparison of screenshot-to-code tools, plugin-based converters, and methodology-driven approaches like Current Labs AI.D for turning designs into code.
-
The Complete Guide to Design-Developer Handoff
Why traditional handoff is broken, common pain points teams face, and how AI.D methodology eliminates the gap between design and development.
-
5 Benefits of AI Development Methodology
Discover how AI.D methodology improves speed, consistency, scalability, and code quality in design-to-code workflows — with concrete before-and-after examples.
-
Introducing AI.D — AI Development Methodology
How AI.D bridges the gap between design and development with a structured, repeatable workflow that turns design systems into production-ready code.
-
Design-to-Code Workflows in 2026 — What's Changed
The landscape of design-to-code tools has evolved rapidly. Here's where things stand and what teams should know about modern workflow options.
-
WebMCP — Making Websites Agent-Friendly
How Current Labs implemented WebMCP on currentlabs.dev to let AI agents interact with the site programmatically, and why it matters for the future of the web.