cd /news/ai-agents/how-are-you-authorizing-ai-agents-th… · home topics ai-agents article
[ARTICLE · art-73655] src=keydris.com ↗ pub= topic=ai-agents verified=true sentiment=· neutral

How are you authorizing AI agents that call MCP servers?

Keydris has launched a trust infrastructure for AI agents that call MCP servers, centered on a Gateway that issues scoped, signed KIT tokens verified by lightweight Kit Reader middleware before any tool call reaches the server. The system provides an append-only, hash-chained audit log and machine-speed revocation across boundaries, designed to meet the standards of CISOs, platform engineers, and compliance leads.

read2 min views1 publishedJul 25, 2026
How are you authorizing AI agents that call MCP servers?
Image: source

Trust Infrastructure for the autonomous economy -

Trust Infrastructure for the autonomous economy -

EVERY ACTION, CHECKED

The Keydris Gateway sits at the heart of the flow: it issues each agent a kit, scoped by the policies you set in the Keydris app. Agents present that kit with every tool call, and the Kit Reader : lightweight middleware installed on top of your MCP server : checks it with the gateway and allows or denies the call before it ever reaches the server.

Verified agent

Kit Reader

doorstep access control

MCP Server

tools & data

Unverified agent

Keydris Gateway

issues agent kits

Keydris App

set policies

WHY THIS EXISTS

The standard of proof

Archetypal roles, not customer quotes — the standard Keydris is built to meet.

“Which of these actions was actually authorized , and by whom?”

Chief information security officer

Every action arrives with its permission attached: issuer, scope, policy, signature , verifiable in milliseconds, independent of the agent that carried it.

issuer acme-corp.keydris.id · sig valid

“If we revoke an agent's authority at 09:00, what does it prove at 09:01?”

Head of platform engineering

Nothing. Revocation is part of verification itself — the next check fails, and the audit trail shows exactly when.

09:00:00Z · authority revoked

09:00:03Z · revocation feed synced

09:00:41Z · action attempted → refused · record signed

“Show me the record you would hand an auditor.”

Compliance lead

A signed sequence of decisions, handed over as it stands: each record names the request, the policy in force, the checks, and the outcome.

record kd_evt_29fa… · signed

THE KIT

And the layer stays neutral. Keydris is independent of any model, vendor, platform, or rail. A requirement for infrastructure that sits between institutions rather than inside one of them.

Signed KIT token

Scoped permissions

Minimum privilege

Keydris issues the agent a signed token, the KIT. It states what the agent may do and for how long. Minimum privilege by default.

Append-only log

Hash-chained

Immutable record

The authorization is bound to an append-only, hash-chained log. An immutable record of what the agent was permitted to do.

Enforced at the front door

Proof checked first

No proof, no connection

Enforcement is at the counterparty's front door. The receiving system verifies the token before the agent can connect or act.

Machine-speed revocation

Across every boundary

Fails verification instantly

Authority is pulled at machine speed, across every boundary. A revoked token fails verification the instant permission should end.

── more in #ai-agents 4 stories · sorted by recency
── more on @keydris 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/how-are-you-authoriz…] indexed:0 read:2min 2026-07-25 ·