Docs

K-Block documentation

Control what your AI agents can do

K-Block sits between your agents and the tools they call. Every call is checked against your organization’s policy, held for a person’s approval when the policy says so, and recorded as a signed receipt that anyone can verify.

Quickstart What is K-Block?

  1. Agent

    calls a tool through the Gateway

  2. Policy

    allow, require approval or deny

  3. Approval

    a person decides, when required

  4. Execution

    the tool runs exactly once

  5. Receipt

    signed evidence of what happened

What happens on every protected tool call (concept)

Explore the docs

Get startedThe architecture, a quickstart and your first protected action.Core conceptsAgents, tools, actions, policies, delegations, approvals and receipts.Using the dashboardWhat each area of K-Block Cloud is for and how to use it.PoliciesWrite rules, test and simulate them, observe, roll out and roll back.Receipts and verificationWhat a receipt proves, and how to verify one yourself.Administration and accessRoles, API keys, audit events, emergency controls and webhooks.SDKsThe TypeScript client, LangChain.js, receipt verification and Python.ReferenceStatuses, error codes, the Cloud API, receipt fields, webhook events and limits.TroubleshootingWhat an error means and how to fix it.