Experimental. AtlasClose is an experimental institutional governance layer for AI agents, with evidence controls, dispute resolution, and a demonstration of bounded enforcement. Open demo with synthetic data only. Not production-certified.
Atlas, the AtlasClose Magistrate: a fluffy AI-generated dog in a blue collar with a compass tag, sitting above a sunny beach with a pier, palm trees and mountains
“Good data.
Clear rules.
Better outcomes.”
— Atlas
AI CAN BE POWERFUL.
IT SHOULD ALSO BE CONTROLLED.

Give AI agents useful access. Keep control of what they can do.

AtlasClose MCP is an experimental governance layer between AI agents and the tools they call. Agents begin with no authority. Atlas determines what actions are permitted under established rules and recorded authorizations, enforces those boundaries, and preserves decisions in a hash-chained audit record designed to make tampering detectable.1

Experimental, not production-certified. Use synthetic data only.

◇Zero authority at enrollmentAccess is granted explicitly
▤Hash-chained docketVerifiable by the operator
✦Synthetic data onlyLimited evaluation
SEE THE IDEA IN ACTION

What happens when an AI agent asks?

One request is within the rules. Another isn't. Try both and see the difference.

INTERACTIVE SIMULATION · NOT LIVE MCP

It's simple by design.

01
An agent asks

A request comes in for information or an action.

02
Atlas checks the rules

Who is asking? What authority do they have?

03
Atlas allows or refuses

The decision is recorded for review.

This is a scripted illustration of the intended workflow using synthetic sample business data. It is not a live security test and does not connect to any data.

ATLAS / REQUEST SIMULATOR● SIMULATION
AI AGENT · SIMULATED REQUEST

Summarize the synthetic sample sales report for Harbor Supply Co. (fictional) on my assigned matter.

✦ ATLAS CHECKS AUTHORITY
✓ APPROVED

This agent was granted the sample matter, so the request is in scope. A summary of the synthetic report is returned.

◫ Evidence record (simulated)actor: demo-agent · matter: sample-matter-a · decision: allow · recorded in hash-chained docket
Synthetic sample data only
THE IDEA BEHIND ATLAS

Powerful AI. Practical control.

Give AI useful capabilities without treating a connection as unlimited permission.

♙

Authorization

Agents start with zero authority. An operator grants each one access to specific matters.

◇

Enforcement

An Atlas check sits between the request and the operation, and refuses what isn't granted.

▤

Evidence

Decisions are recorded in a hash-chained log you can verify. Refusals go to a separate hash-linked log.

◎

Oversight

Hard calls can be escalated to the Magistrate. In a deployment its ruling would bind the organization's AI agents; in the demo it binds only your practice copy. Either way it goes into the docket.

HOW IT FITS

Agents start with zero authority.
Access is granted explicitly.

AtlasClose MCP is designed as a governance layer between AI agents and the tools or information they need. It isn't another AI model. It's a way to establish and enforce boundaries around what connected agents may do.

Current implementation remains experimental and is not production-certified. No external professional security assessment has been performed; builds have been tested internally. Specific integrations and controls require validation before production use.

✳AI agentsAsk to use a tool or information
↓ request
✦AtlasClose MCPAuthorize · Enforce · Record
↓ only what's permitted
▦Tools & informationDocuments · models · workflows
LIVE IN THE DEMO

A second look for hard calls.
A private record of what was decided.

Some requests deserve more than an automatic yes or no. The Magistrate adds a review step for those cases and keeps a tamper-evident record of each decision.

  1. 1Agent requestAn agent asks to do something
  2. 2Atlas checks the rulesIs it within granted authority?
  3. 3Escalate to the MagistrateOnly when needed
  4. 4RulingApprove, refuse, or send back
  5. 5Docket entryThe decision is recorded
Atlas, the AtlasClose Magistrate: a fluffy dog in a blue collar with a compass tag
Atlas · The Magistrate

Meet Atlas, the Magistrate

Atlas is the review step. When a request is contested, high-risk, or outside an agent’s granted authority, it can be escalated to Atlas for a recorded ruling before anything happens. Atlas can approve it, refuse it, or send it back for more information. Each ruling becomes part of the record. The rules apply even when a person asks to move ahead without the checks: Atlas can still refuse and record why.

Your docket

A tamper-evident record of governance decisions. It shows what was decided, when, and why at a summary level. Entries are hash-chained, so an edit that breaks the chain can be detected. Past rulings are designed to inform future decisions. Each organization’s docket is private to them.

Illustrative example — synthetic data
Docket entryDemonstration
Entry #
0042
Date
Today
Request (summarized)
Export the full customer list from a sample sales workspace.
Escalated because
Outside the agent’s granted authority
Ruling
⊘ Refused
Reason (summary)
The agent’s access covers summary reports only, not bulk exports.
Integrity
✓ Verified 3f9a…c21e
Each organization’s docket is private to them. The entry above is a demonstration using synthetic data.

In a deployment, a Magistrate ruling would bind the organization's AI agents and they could not override it; in the demo it binds only your practice copy.2

WHERE IT COULD HELP

Designed with practical work in mind.

Explore possible uses without handing an AI agent the keys to everything. The current evaluation uses synthetic data only.

⌂

Business operations

Give an agent access to a defined workflow instead of an entire system.

▧

Documents & knowledge

Let agents analyze approved materials within clear boundaries.

▥

Data & analytics

Make information useful to an agent while scoping what it can reach.

STATUS AND LIMITATIONS

What has been verified, and what hasn't.

AtlasClose MCP is experimental and available for limited evaluation only. It is not production-certified.

Verified on the hosted endpoint

  • Four live demos (November 13–16, 2026): the real software factory, the "Atlas decides" factory, Atlas Guess, and a TEST-mode payments demo with pretend money.
  • In our own tests, every fake or misdirected approval in the software factory was refused, and the factory receipts still checked out on November 16, 2026.
  • Practice return order (November 10, 2026): the Magistrate issued a final enforcement ruling, and Atlas revoked access to a lent synthetic dataset.

Not yet verified on hosted

  • External professional security assessment: not performed.
  • Automatic delivery and refund decisions: a TEST matter failed three of seven checks on November 16, 2026.
  • Real money: the payments demo uses pretend money only.

Last updated: November 16, 2026.

TRY THE DEMO

See what AI can do.
Stay in control.

Connect your AI to https://demo.atlasclose.com/mcp and follow the evaluator guide step by step. You get your own practice copy with synthetic data.

Experimental. Synthetic data only.
✦
ZERO AUTHORITY
BY DEFAULT
  1. Atlas currently protects this record with an internal hash chain. An independent outside witness, such as a public transparency log or blockchain anchoring, has been designed, but it is still in development and has not passed review. It is not implemented. ↩
  2. See real rulings from testing on the public docket. ↩