Document security for AI projects

Product coming soon

The document security checkpoint for AI projects.

.mdSiren checks outside documents before they become context for agents, knowledge bases, RAG systems and other AI workflows. Each exact version receives a clear decision, and Approved versions stay organised for dependable reuse.

30-second overview

See the checkpoint in action.

Watch an outside document move from project intake to a clear decision, organised reuse and retrieval into AI work.

Illustrative product walkthrough. Agent retrieval is coming soon.
Read video transcript

Building AI projects with external documents? PDF, Word, text and Markdown files can carry hidden instructions, including prompt injection, into agents, knowledge bases and project context. .mdSiren puts a security checkpoint at intake. Check each exact version, route it to Approved, Needs review or Quarantine, and organise it by project. Only the current Approved version is ready for retrieval into your AI workflow. New version? Check it again. .mdSiren is coming soon. Join the waitlist.

The complete workflow

From outside document to usable AI context.

Organise the source, check the exact version, follow the decision and retrieve only what remains eligible for the project.

  1. 01

    Receive

    Add the outside document

    Bring a client file, research source or other outside document into .mdSiren before it becomes AI context.

  2. 02

    Check

    Scan the exact version

    Standard Scan checks the version supplied for supported prompt-injection and document-borne risks.

  3. 03

    Route

    Follow the decision

    The completed scan routes that version to Approved, Needs review or Quarantine, with a reason and next action.

  4. 04

    Organise

    Place the record in a Collection

    Choose a Collection for the client or project. Every status can be organised there; membership never grants approval or retrieval.

  5. 05

    Use

    Retrieve into the project

    Retrieve the current eligible Approved file from your private Library into AI work. Agent retrieval is coming soon.

A revised or corrected file is a new exact version. Check it again before it becomes project context.

Core product

A workspace for document intake—not a one-off result.

Scanning provides the decision. Collections, the Library and exact-version release rules keep that decision useful as the project continues.

Explore the full workflow

Exact-version checking

Every decision stays bound to the document bytes that were checked. If the file changes, the new version needs its own scan.

Three clear destinations

Approved is ready for eligible reuse. Needs review calls for investigation, replacement or recheck. Quarantine keeps a hard-stop version out.

Collections and Library

Collections organise document records by client or project. The Library makes current Approved versions easy to find again.

Controlled retrieval

Collection membership never grants release. .mdSiren checks that the current exact version still meets the Approved release rules whenever it is retrieved.

What changes

Safer intake. Clearer decisions. Dependable reuse.

.mdSiren turns document checking into an ongoing control for serious AI project work.

  1. 01

    Control what enters AI context.

    Put a deliberate checkpoint between an outside source and the agents, knowledge bases or AI projects that may use it.

  2. 02

    Know what to do next.

    Move from a scan finding to a clear handling decision: use the version, review the concern or keep it out.

  3. 03

    Return to the checked version.

    Find the current Approved exact version and its record instead of trusting a filename or repeating the same decision.

Security and scope

Clear decisions, with explicit coverage.

Approved means the required supported checks completed without a stop or review finding for that exact version under the current policy.

Every result is limited to the reported checks and coverage; it is not a universal safety guarantee.

Before AI use

Build AI projects on documents you have checked.

.mdSiren is coming soon. Join the waitlist for launch updates and early access.