Code. Captain. Trade.
One control plane for AI work
Gouvernail routes intent to product engines, scopes context, runs agents in bounded environments, and asks the human only when judgment is required.
Built and Shipping
Code is live by default. Captain, Trade, and multi-model ship behind feature flags.
Product routing
One account routes work to Code, Captain, or Trade from subdomain, project, or phone number.
Shared work units
Every delegated job has state, ownership, triggers, audit events, and a product-specific engine.
WhatsApp and SMS control
Create work, approve decisions, rate findings, and respond to proposals from the phone.
Product-aware dashboard
The same web shell renders Code projects, Captain research vaults, and Trade workspaces.
HITL triggers
Gouvernail asks for judgment on risky diffs, stuck agents, findings, and approval-gated proposals.
Audit journal
Work-unit state, decisions, provider metadata, research findings, and proposal outcomes are recorded.
Code sandboxes
Coding agents run headlessly in bounded workspaces with git, PR, preview, merge, and rollback flows.
Git workflow
Clone, branch, commit, push, open PRs, deploy previews, and merge with human approval when needed.
Captain findings
Push links, brainstorm, review findings, rate signal quality, and keep a private knowledge graph.
Captain settings
Privacy mode, cross-product context opt-in, cadence, and model preference are project settings.
Trade committee
Research portfolios and watchlists with committee rationale, sources, risks, and required disclaimers.
Approval-first trading
Paper trading is the default, and every execution-like proposal waits for explicit approval.
Model selection
Claude remains the default while model routing can use authorized OpenAI-compatible providers.
OSS sidecar support
Self-hosted models such as Scaleway Qwen can be exposed through the OpenAI-compatible local provider.
Feature-flag rollout
Captain, Trade, and multi-model can ship dark, canary safely, and roll back independently.
Production runbooks
Env vars, migrations, canary steps, rollback plans, and monitoring checks are documented per product.
Next Product Gaps
These are the launch-readiness items after the multi-product pivot.
Full portfolio canary
Run Code, Captain, Trade, and model selection together with real Twilio, GitHub, DB, web, worker, vault, and Scaleway dependencies.
Portfolio onboarding
Product-specific first-run flows for a Code PR, a Captain vault, and a Trade watchlist.
Provider scorecards
Expose model/provider outcomes, fallback behavior, cost, and satisfaction by task class.
Admin policy
Enterprise controls for enabled products, provider allowlists, context-sharing policy, and audit export.
Start with the product that matches the work
Same account, same phone login, product-specific workspace.