Overview
What Verybear is, how a request becomes a flow, and where to start.
Verybear is a task router for company AI usage. People keep using the AI tools they already have (Claude, Codex, Pi, Paseo, Cursor, VS Code) and type plain requests like they always have. Verybear works out what the task actually is, then routes it into a flow your company defined: a model, an effort level, a playbook, and sometimes a subagent breakdown.
Nobody picks a model. Nobody holds an API key. Nobody sees the instructions that get injected before the model runs. They just get the result your best person would have produced.
The three pieces
A desk is a person's connected tool: Priya's Pi, Tom's Codex, Sofia's Paseo. A flow is a defined task with a model, effort level, and playbook attached. Policy is the org-wide layer on top: approved models, budgets, blocked request types, set once and synced to every desk.
- Your AI team defines flows and policy in the console or with the CLI.
- Verybear syncs that configuration to every connected desk, automatically.
- Anyone on the team just asks for what they need, in whatever tool they're already in.
What happens to requests you haven't defined a flow for
They land on a safe default, General assistance, get logged, and show up later as a proposed new flow once a real pattern shows up. Nothing falls through unhandled, and nothing gets silently guessed at.