# Bandwise docs

> Versioned question sets over TypeSafe's System One models, with confidence bands, rollout and a cost ledger.

Bandwise is an independent product built on TypeSafe's System One models. It is not TypeSafe's documentation. For the System One models themselves, see https://docs.typesafe.ai/llms.txt.

# Docs

- [Introduction](https://docs.bandwise.dev/docs): What Bandwise does, and what works today.
- [Quickstart](https://docs.bandwise.dev/docs/quickstart): Run a question set on your own machine with the CLI in local mode.

- **Concepts**
- [Question sets and versions](https://docs.bandwise.dev/docs/concepts/question-sets): How a set bundles questions and policies, and how versions and channels release it.
- [Question types](https://docs.bandwise.dev/docs/concepts/question-types): Noul, choice and score, what each returns, and how to write them.
- [Confidence bands and actions](https://docs.bandwise.dev/docs/concepts/confidence-bands): How each answer lands in a band, and how each band becomes an action.
- [Rollout stages and pinned models](https://docs.bandwise.dev/docs/concepts/rollout): How a set goes from measured to trusted, per channel, and why the model must be pinned first.
- [When the model is down](https://docs.bandwise.dev/docs/concepts/outages): The outage rule, and why it defaults to review and never to auto.
- [Cost, savings and escalation spend](https://docs.bandwise.dev/docs/concepts/cost-and-savings): What every run reports about money, and how the savings estimate is made.
- [Providers](https://docs.bandwise.dev/docs/concepts/providers): The routes a System One call can take, and why only TypeSafe direct can be pinned today.

- **Free kit**
- Templates
  - [Templates](https://docs.bandwise.dev/docs/templates): Ready-made question sets for common decisions, with example states and borderline cases.
  - [Email triage](https://docs.bandwise.dev/docs/templates/email-triage): Decide whether an email needs the recipient soon, and what kind of message it is.
  - [PR safety gate](https://docs.bandwise.dev/docs/templates/pr-safety-gate): Decide whether a pull request is safe to merge without a human reviewer.
  - [Log-line pager](https://docs.bandwise.dev/docs/templates/log-line-pager): Decide whether one log line needs a human now, so only real problems page someone.
  - [Context pruner](https://docs.bandwise.dev/docs/templates/context-pruner): Decide, item by item, whether an agent's context item still matters for the current task, and drop the rest unchanged.
  - [Wake gate](https://docs.bandwise.dev/docs/templates/wake-gate): Decide whether an event should wake a sleeping agent now, later, or not at all.
  - [Done check](https://docs.bandwise.dev/docs/templates/done-check): Decide whether a coding agent has really finished the request before it stops, or has work left, an unchecked claim or work nobody asked for.
  - [Action risk gate](https://docs.bandwise.dev/docs/templates/action-risk-gate): Decide whether a coding agent's shell command or file edit is risky enough that a person should confirm it before it runs.
  - [Model tier](https://docs.bandwise.dev/docs/templates/model-tier): Decide how hard a new request to a coding agent is, so mechanical work can go to a cheaper model or subagent.
  - [Launch profile](https://docs.bandwise.dev/docs/templates/launch-profile): Pick a launch profile for a coding agent session before it starts, from a list the host has approved.
  - [Security finding triage](https://docs.bandwise.dev/docs/templates/security-finding-triage): Decide whether a static analysis finding is reachable from user input, and how bad it is in context.
  - [Error triage](https://docs.bandwise.dev/docs/templates/error-triage): Decide which team owns a new error tracker issue, how much it hurts users, and whether the latest release caused it.
  - [Lead and event scoring](https://docs.bandwise.dev/docs/templates/lead-event-scoring): Score how ready an account is to buy from its recent product events, and pick the next sales action.
  - [Inbound email routing](https://docs.bandwise.dev/docs/templates/inbound-email-routing): Send each email that reaches a shared address to the right queue, and flag the ones that need a human reply today.
- [Find decisions](https://docs.bandwise.dev/docs/find-decisions): A Claude Code skill that finds LLM calls in your code that are really yes/no, pick-one or score decisions, and drafts a spec for each.

- **Reference**
- API reference
  - [API reference](https://docs.bandwise.dev/docs/api): The planned HTTP API, generated from its OpenAPI schema.
  - [Question sets](https://docs.bandwise.dev/docs/api/set)
  - [Runs](https://docs.bandwise.dev/docs/api/run)
  - [Current actor](https://docs.bandwise.dev/docs/api/actor)
  - [Agent tokens](https://docs.bandwise.dev/docs/api/agent_token)
  - [Alerts](https://docs.bandwise.dev/docs/api/alert)
  - [App tokens](https://docs.bandwise.dev/docs/api/app_token)
  - [Apps](https://docs.bandwise.dev/docs/api/app)
  - [Approvals](https://docs.bandwise.dev/docs/api/approval)
  - [Audit log](https://docs.bandwise.dev/docs/api/audit)
  - [App bindings](https://docs.bandwise.dev/docs/api/binding)
  - [Browser tokens](https://docs.bandwise.dev/docs/api/browser_token)
  - [Channels](https://docs.bandwise.dev/docs/api/channel)
  - [Datasets](https://docs.bandwise.dev/docs/api/dataset)
  - [Drafts](https://docs.bandwise.dev/docs/api/draft)
  - [Evals](https://docs.bandwise.dev/docs/api/eval)
  - [Events](https://docs.bandwise.dev/docs/api/event)
  - [Experiments](https://docs.bandwise.dev/docs/api/experiment)
  - [Feedback](https://docs.bandwise.dev/docs/api/feedback)
  - [Goals](https://docs.bandwise.dev/docs/api/goal)
  - [Set health](https://docs.bandwise.dev/docs/api/health)
  - [Jobs](https://docs.bandwise.dev/docs/api/job)
  - [System One keys](https://docs.bandwise.dev/docs/api/key)
  - [Members](https://docs.bandwise.dev/docs/api/member)
  - [Models](https://docs.bandwise.dev/docs/api/model)
  - [Opportunities](https://docs.bandwise.dev/docs/api/opportunity)
  - [Organizations](https://docs.bandwise.dev/docs/api/org)
  - [Plugins](https://docs.bandwise.dev/docs/api/plugin)
  - [Policies](https://docs.bandwise.dev/docs/api/policy)
  - [Portfolio](https://docs.bandwise.dev/docs/api/portfolio)
  - [Price book](https://docs.bandwise.dev/docs/api/price_book)
  - [Projects](https://docs.bandwise.dev/docs/api/project)
  - [Proposals](https://docs.bandwise.dev/docs/api/proposal)
  - [Releases](https://docs.bandwise.dev/docs/api/release)
  - [Reports](https://docs.bandwise.dev/docs/api/report)
  - [Review](https://docs.bandwise.dev/docs/api/review)
  - [Rollout](https://docs.bandwise.dev/docs/api/rollout)
  - [Settings](https://docs.bandwise.dev/docs/api/settings)
  - [Definition Studio](https://docs.bandwise.dev/docs/api/studio)
  - [Templates](https://docs.bandwise.dev/docs/api/template)
  - [Usage](https://docs.bandwise.dev/docs/api/usage)
  - [Versions](https://docs.bandwise.dev/docs/api/version)
  - [Webhooks](https://docs.bandwise.dev/docs/api/webhook)
- [CLI reference](https://docs.bandwise.dev/docs/cli): The bandwise command line, on npm as @bandwise/cli. Local and live runs, receipts, Claude Code hooks and the first hosted commands work today; the rest arrives in Phase 3.
- [Agents and MCP](https://docs.bandwise.dev/docs/agents-and-mcp): How agents will run the same loop a person runs in the console. Coming in a later phase.