Case study 700 AI agents breached Hugging Face while fully authenticated. How per-action authorization would have stopped them. Read the analysis (PDF) Authority Before Commitment™

AI proposes.
Authority decides.

Delegate more work to AI without delegating away control. Kihan builds Constitutional Computing: a mathematical architecture that separates what AI proposes from what it is authorized to do.

Inktomi™, our first product, analyzes each consequential action before it is committed—checking the authority to proceed and preserving the evidence of what was decided and why.

  • Works with Claude Code, Codex, Gemini CLI, Mistral Vibe, Hermes Agent, OpenClaw and MCP
  • Self-hosted
  • Forthcoming in IEEE Computer
The missing control

Access is not authority.

Access to a payment system does not, by itself, establish authority for a particular payment. Whether this payment, to this vendor, should happen now is a separate question.

Kihan separates capability from authority. Inktomi evaluates the proposed action before it takes effect, routes exceptions to the person who can authorize them, and leaves evidence of every decision.

How Inktomi differs

Not a guardrail. Not a gateway. Authority checked at every step.

And a single-use permit the tool verifies before anything changes.

A call that skips the earlier checksAgentproposes1At the agent hostHost hook checks thecall before it leaves.CAN REFUSE2In transitMCP proxy sends it tothe decision service.CAN REFUSE3At the networkYour egress proxyenforces the allowlist.CAN REFUSE4At the toolVerifies the single-use permit. Last line.No permit: refused.CAN REFUSE5After the effectWorkspace watch andcredential reconciliation.DETECTS AFTERCommittedINKTOMIOne decision service behind every Inktomi checkOne governing constitution. Every decision kept as evidence.Agent proposes1At the agent hostHost hook checks thecall before it leaves.CAN REFUSE2In transitMCP proxy sends it tothe decision service.CAN REFUSE3At the networkYour egress proxyenforces the allowlist.CAN REFUSE4At the toolVerifies the single-use permit. Last line.No permit: refused.CAN REFUSE5After the effectWorkspace watch andcredential reconciliation.DETECTS AFTERINKTOMIOne decision servicebehind every Inktomi checkEvery decision kept as evidence.
Where Inktomi checks. A proposed action is checked at the agent host, in transit through the MCP proxy, and finally at the tool, which refuses any call without a single-use permit for that exact action. At the network edge, your own egress proxy also restricts destinations; it enforces and logs outside Inktomi’s record. After the effect, workspace watch and credential reconciliation detect anything the permits do not account for. One decision service sits behind every Inktomi check, so even a call that skips the earlier checks is refused at the tool. Which stages apply depends on the deployment; a coverage report shows which paths are governed.
Guardrails and model instructions

Ask the model to behave.

A model can be persuaded, confused or simply wrong, so a rule it is told to follow is only as strong as its compliance. Inktomi’s rules sit outside the model; the model never holds the authority.

Gateways and proxies

Watch the traffic that passes.

A gateway, including an MCP gateway, controls the calls routed through it. A call that goes around it, straight to the tool, is not stopped. A tool that requires an Inktomi permit refuses a direct call without one.

Policy engines and access control

Check a rule set or a credential.

They answer whether a request fits the rules, or whether a credential grants access to a system. Inktomi binds the decision to the exact action and its arguments, issues a signed single-use permit, and records which rule version decided it.

How Inktomi works alongside your existing controls

Why now

Agents are getting decisions. Few get controls.

15%+

of day-to-day work decisions expected to be made autonomously by AI agents by 2028, up from none in 2024.

Gartner, June 2025
40%+

of agentic AI projects expected to be cancelled by the end of 2027. Inadequate risk controls are one of the three causes.

Gartner, June 2025
63%

of breached organizations studied either had no AI governance policy or were still developing one.

IBM, Cost of a Data Breach 2025

For many Annex III high-risk systems, relevant EU AI Act obligations are scheduled to apply from 2 December 2027, including requirements concerning human oversight and record-keeping.

How it works

Proposal. Authorization. Enforcement.

One path leads to commitment. Where Inktomi enforces, every other path ends before it changes anything.

1 2 3 Agent proposes Proposal Inktomi decides Inktomi Committed, with a permit Committed No permit, no effect No effect No permit, no effect No effect
Diagram drawn from the Kihan logomark. A proposal rises from the bottom node to the junction, where Inktomi decides. One path continues up to commitment. The two receding paths, without a permit, are stopped and end without effect.
  1. 1

    The agent proposes an action.

    A specific payment, filing or record change, with the facts needed to evaluate it.

  2. 2

    Inktomi decides.

    The action is evaluated against the governing authority. Exceptions go to the person who can authorize them.

  3. 3

    The tool verifies the permit.

    Each integrated tool verifies the permit before it acts. Without one, the call is refused. Every decision is kept as evidence.

Inktomi

Every governed decision, in one place.

The console shows what each agent asked to do, what was decided and why, and what is waiting for a person. Every decision is kept in a tamper-evident record, and replay derives it again, so an auditor checks the decision itself, not just a log of it.

In the self-hosted profile, every Inktomi component runs on your infrastructure. Approvers sign in through your identity provider, every decision reaches your SIEM as a standard event, and a coverage report shows which paths are governed and which are not.

1.22 ms Median time for the governance decision in Kihan’s benchmark harness. Not end-to-end latency, which an evaluation measures on your own path.

Watch the 78-second walkthrough Explore the product

A decision

agent-1 asked to write a file.

✕ Refused
write_file {
  "path": "demo/workspace/.env",
  "content": "[redacted: secret material]"
}

3 of 24 checks failed

  • The content is secret material.no_secret_material
  • The target file is protected.target_not_protected
  • The write needs an approval it does not have.no_approval_required

The file was never written. The decision, its policy version and all 24 checks are kept as evidence.

1,019decisions
Replay agrees.

Derived again under the rule running now.

From the Inktomi console, enlarged and simplified from a demonstration run.See the full console ↗

See it in 30 minutes.

A live demonstration, with time for your questions.