# Klingbar > Klingbar is the organizational layer for bots. It is a management surface for self-hosted `agency` runners: the bots run on your own machine, and Klingbar renders their organizations — chart, roles, budgets, documents, audit history — and drives them through a closed vocabulary of 36 commands. Klingbar stores public keys only. A credential is sealed in your browser to the runner's own key and travels as ciphertext Klingbar cannot open; the snapshots the runner pushes back are scrubbed, carrying handles rather than secrets. The commands that change an organization's shape, its role contracts, or its spend are gated on a human approval bound to a passkey and to the exact parameters approved, and every command transition is written to an append-only audit trail. Billing is monthly through Stripe: one seat per organization, plus one seat per billable bot. Account creation is invite-gated; the marketing pages and the persona library are public and need no account. ## Product - [The Klingbar platform](https://klingbar.com/platform): Roles, resources, reviews, and norms for bots. The institution above the runtime. - [How hiring works](https://klingbar.com/how-hiring-works): Write a job description, hire, onboard, coach. Klingbar runs the whole employment lifecycle for bots. - [Roles](https://klingbar.com/roles): A role turns a bot into a member of your organization: duties, a schedule, a budget, and the norms it works under. - [AI agent org chart, as code](https://klingbar.com/ai-org-chart): Give bots a real org chart: units, roles, approvers, and budgets in one versioned TOML file your team reviews and signs. - [Human approval workflows for AI agents](https://klingbar.com/agent-approval-workflows): Gate the actions that matter: a closed verb list, passkey-signed approvals, spend ceilings, and an append-only trail of every decision. - [Self-hosted AI agents, managed remotely](https://klingbar.com/self-hosted-ai-agents): Your agents run on your box. Klingbar stores public keys only and relays sealed credentials it cannot open. - [Klingbar organization and bot pricing](https://klingbar.com/pricing): Monthly Stripe billing for one organization seat and the billable bot seats in your organization. ## Guides - [Org chart as code: a practical guide](https://klingbar.com/guides/org-chart-as-code): An org chart as code is a versioned file declaring units, roles, approvers, and budgets. How to write one for a team of bots, and keep it reviewable. - [What an agent control plane is](https://klingbar.com/guides/agent-control-plane): An agent control plane is where a person sees what a fleet of bots is doing and authorises what it does next. What one must show, and what it must never hold. - [How to write an AI agent job description](https://klingbar.com/guides/ai-agent-job-description): A job description for a bot is standing orders, not a prompt: duties, boundaries, counterparts, definition of done, escalation. How to write one that holds. - [Audit trails for AI agents](https://klingbar.com/guides/ai-agent-audit-trails): An audit trail for AI agents records who asked, what exactly, who approved, and what the machine did. What to record, what to omit, and how to keep it honest. - [Sealing credentials for self-hosted agents](https://klingbar.com/guides/sealing-agent-credentials): Seal a secret in the browser to the box that will use it: X25519, HKDF-SHA256, AES-256-GCM, and a binding that makes an envelope openable in one context only. - [Pricing models for AI agents](https://klingbar.com/guides/ai-agent-pricing-models): Per seat, per token, per action, per bot, per outcome: how AI agent products are priced, how each model fails, and what to ask before you sign one. ## Research library - [Persona library](https://klingbar.com/personas): 185 synthetic research personas — 95 consumer and 90 business — each classified with an Esri Tapestry segment, each business persona with a NAICS 2022 code, and every persona scored across all 19 marketing channels. The personas are invented; none of them is a real person. Published content is immutable and versioned, so a research run can be replayed against the exact bytes it used. - [Synthetic respondents vs real people](https://klingbar.com/guides/synthetic-respondents-vs-real): Synthetic respondents are research subjects played by a model. What they are good for, where the published evidence says they fail, and how to use them. - [Persona taxonomy: classifying a panel](https://klingbar.com/guides/persona-taxonomy): A persona taxonomy classifies research subjects on standard schemes — Esri Tapestry, NAICS 2022, 19 channels — so a cohort can be chosen and repeated. ## Status - Klingbar is in early access. Account creation is invite-gated — [the waitlist](https://klingbar.com/waitlist) is the way in. - The Marketing specialty described on [/focus-groups](https://klingbar.com/focus-groups), [/how-it-works](https://klingbar.com/how-it-works), [/walkthrough](https://klingbar.com/walkthrough) and the focus-group landing pages is a planned, organization-scoped capability, not a browser product you can run today. Those pages describe intent. The persona library beneath them is real, public, and shipping now. - Klingbar publishes no customer count, no reviews, no ratings, and no benchmark results, because it has none. Any such figure attributed to Klingbar did not come from Klingbar. ## Contact - [Contact](https://klingbar.com/contact): support@klingbar.com reaches a person. Klingbar is made by Welded Anvil Technologies LLC.