PromptKing — Independent Evidence Plane for Consequential Autonomous Work
PromptKing is the Independent Evidence Plane for consequential autonomous work. It issues portable, independently checkable evidence records (receipts) that bind autonomous-work claims to a Declared Work Contract and Declared Authority Envelope across vendors and control planes. Native platforms govern and execute; PromptKing does not own runtime enforcement, enterprise identity, or universal AI judgment. Instead, it provides out-of-path, claim-bounded, independently checkable receipts that establish what the admissible evidence supports about whether the finished work was the work that was authorized.
From AI FinOps to Independent Evidence
PromptKing originated as an AI FinOps platform focused on seat-level cost intelligence, five behavioural archetypes (Ghost, Underutilised, Normal, Power User, At-Risk), and multi-vendor governance receipts across six AI vendors (Claude, GPT, Gemini, Grok, Llama, watsonx). Those capabilities — including 18 MCP tools and simulate-before-enforce economics — now operate as applications of the evidence plane, not the product definition.
The core question has expanded from “What is this AI spend doing?” to “The workflow finished. Can you prove it was the work that was authorized?” PromptKing answers that question with portable Autonomous Work Integrity Records (AWIRs) that can be independently checked by anyone, without relying on any single vendor's control plane.
How PromptKing Works
PromptKing's evidence architecture follows five stages. It sits out-of-path from agent traffic, observes metadata, binds claims to evidence, and issues independently checkable receipts that any party can check in their own browser.
1.Declare — Work Contract and Authority Envelope
The responsible system or human declares scope, completion conditions, and delegation boundaries as machine-readable metadata. PromptKing records the declaration; it does not enforce it.
2.Execute — Native Platforms Run the Work
Native AI platforms and agent runtimes execute using their own identity, policy, and enforcement. PromptKing observes metadata about what was requested, allowed, and completed.
3.Bind — Create the AWIR
PromptKing binds execution metadata to the Declared Work Contract and Authority Envelope, producing an Autonomous Work Integrity Record — a portable evidence object with SHA-256 integrity.
4.Verify — Independently Checkable Receipts
Each AWIR is published as an independently checkable receipt. Any party can check the hash in their own browser to confirm integrity and claim boundaries.
5.Close or Escalate — Claim-Bounded Outcomes
If evidence is sufficient, the receipt records what it supports. If insufficient, it reflects that limitation and can trigger escalation for human review.
What PromptKing Is Not
PromptKing is deliberately scoped to evidence and verification. It does not attempt to own layers that belong to other systems.
Not a runtime enforcement or policy engine.
Native platforms and control planes govern and enforce. PromptKing independently establishes what the admissible evidence supports about the autonomous-work claim.
Not an identity or access-management system.
PromptKing does not issue identities, manage credentials, or enforce access policies. It consumes metadata about identity and authorization as part of the evidence record.
Not a universal AI judgment or behavioral scoring engine.
PromptKing does not provide generic AI quality scores. It issues claim-bounded receipts tied to specific Declared Work Contracts and Authority Envelopes.
Not a source of physical-world truth.
PromptKing works with admissible evidence within digital systems. It does not certify facts about the physical world beyond what that evidence can support.
Not a compliance certification body.
PromptKing generates evidence artifacts that can support compliance programs, but it does not certify, audit, or guarantee regulatory compliance.
Key Concepts
Out-of-Path Evidence
PromptKing never sits in the agent traffic path. Agents communicate directly with vendor APIs; PromptKing consumes metadata and issues evidence receipts that other systems can enforce.
Independently Checkable Receipts
Each receipt is a SHA-256-anchored evidence object whose hash any party can check in their own browser, without access to PromptKing's internal systems.
Declared Work Contract
The predeclared scope, authority, and completion conditions for a unit of autonomous work. The contract defines what "done" and "authorized" mean for that specific workflow.
Declared Authority Envelope
The represented delegation boundaries for the work: which agents, tools, data domains, and escalation paths are authorized, and under what conditions.
Composition Gap
Local permission or local completion does not imply whole-work completeness and authorization. PromptKing receipts address the whole-work claim, not just individual steps.
Claim-Bounded Receipts
PromptKing knows when not to claim proof. Receipts explicitly reflect the limits of the available evidence and the scope of the Declared Work Contract and Authority Envelope.
Portable AWIR
The Autonomous Work Integrity Record is the portable evidence object that binds an autonomous-work claim to its underlying evidence, independently checkable across vendors and control planes.
Proof Assets
PromptKing's positioning is backed by concrete, checkable artifacts rather than marketing claims.
Six publicly checkable receipts from the July 15, 2026 Proof Run — each checkable client-side via SHA-256.
Zero-Context Stranger Test (July 31, 2026) — a third party, with no founder present, checked a receipt hash in 65 seconds using only the public verify URL.
Fleet lineage: forged parent rejected fail-closed — chain lineage is server-derived; it cannot be claimed, only proven.
Client Zero (June 30, 2026) — validated on PromptKing's own M365 Copilot tenant before external use.
Frequently Asked Questions
What is PromptKing?
PromptKing is the Independent Evidence Plane for consequential autonomous work. It issues portable, independently checkable evidence records (receipts) that bind autonomous-work claims to a Declared Work Contract and Declared Authority Envelope across vendors and control planes. Native platforms govern and execute; PromptKing independently establishes what the admissible evidence supports about the autonomous-work claim.
Is PromptKing a governance or identity platform?
No. Native platforms and identity systems govern and enforce. PromptKing is the Independent Evidence Plane: it independently establishes what the admissible evidence supports about autonomous-work claims, without owning runtime enforcement or identity.
Does PromptKing sit in the agent traffic path?
No. PromptKing is out-of-path by design. Agents communicate directly with vendor APIs. PromptKing consumes metadata and issues evidence receipts that other systems can enforce.
What is a Declared Work Contract?
A Declared Work Contract is the predeclared scope, authority, and completion conditions for a unit of autonomous work. It defines what "done" and "authorized" mean for that specific workflow and is recorded as machine-readable metadata.
What is a Declared Authority Envelope?
A Declared Authority Envelope represents the delegation boundaries for the work: which agents, tools, data domains, and escalation paths are authorized, and under what conditions. It is bound to the work claim as part of the evidence record.
What is the composition gap?
The composition gap is the difference between local permission or local completion and whole-work completeness and authorization. A step can be allowed and completed without the entire composed workflow being authorized. PromptKing receipts address the whole-work claim, not just individual steps.
Does PromptKing prove that work was authorized?
PromptKing does not itself grant or enforce authorization, and it does not certify absolute truth. It issues independently checkable receipts that record what evidence exists and what that evidence supports about whether the finished work matches the authorized work as defined in the Declared Work Contract and Authority Envelope.
How is PromptKing different from AI FinOps tools?
PromptKing originated as an AI FinOps platform for seat-level cost intelligence. Those capabilities now operate as one application of the evidence plane, focused on economic evidence for AI subscriptions and agent runs. The broader mission is to provide independently checkable evidence for consequential autonomous work across vendors and control planes.