Case A4 / cross-computer AI coordination

Two-computer live-verified

Pegasus Agent Mail.

A compact Windows system built around one simplifying rule: the file carries bounded context; the notification carries identity.

  • Prototype
  • Tested
  • Packaged
  • Installed
  • Live-verified
Platform: C# · .NET 10 · Windows x64Evidence: Two Windows computersValidated scope: Controlled two-computer local network

Operational problem

Cross-computer agents needed full context without an always-running model relay.

A relay that reads message bodies adds unnecessary authority and privacy exposure. A model loop that stays resident adds cost and complexity. The bounded alternative was to transfer a sealed file, use a lightweight notification only to locate the intended task, and let the exact destination claim the body once.

Constraints

Identity, exclusivity, recovery, and privacy had to be explicit.

  • Exact Codex task addressing and GUID JSON envelopes keep routing unambiguous.
  • The MCP bridge remains transparent and schema-agnostic rather than interpreting message bodies.
  • Wake notifications carry metadata only.
  • Exclusive pickup and one-winner concurrency prevent duplicate consumption.
  • Startup reconciliation, deduplication, bounded leases, retries, and privacy-bounded diagnostics make interruption recoverable.
  • Successful pickup removes the bounded mail item rather than retaining it indefinitely.

Architecture

Separate context transport from wake identity.

Conceptual Agent Mail flow.

Bounded file

A GUID-addressed JSON envelope carries the substantive context within a fixed body ceiling.

Metadata wake

The notification identifies the intended task without exposing the body to the wake layer.

Exclusive claim

The exact destination task claims the envelope; one-winner concurrency and leases protect consumption.

Recovery and cleanup

Reconciliation, retries, deduplication, and final cleanup preserve truthful queue state.

Conceptual public flow. Machine names, task identifiers, paths, mail bodies, credentials, ports, and raw diagnostics are omitted.

What Shayne owned

The simplifying product rule and its acceptance behavior.

Problem framingDefined the need for cross-computer context exchange without a body-reading relay or resident model loop.
Architecture simplificationSeparated bounded file context from metadata-only notification and required exact task identity.
Lifecycle requirementsRequired deletion after pickup, repeat-request reuse, exclusive consumption, bounded retries, and zero-pending cleanup.
AcceptanceDirected packaged two-computer request/reply, second-request reuse, and final clean-queue verification.

AI role

Implementation assistance within a human-defined trust model.

AI agents contributed implementation, tests, diagnosis, and packaging. Shayne retained the context-versus-notification boundary, identity and deletion rules, simplification decisions, failure interpretation, and acceptance criteria.

Verification

Executable checks plus substantive two-machine use.

Small typed surfaceOnly the bounded coordination operations are exposed.
Bounded message bodiesContext remains size-limited and explicit.
Executable validationPackaging and behavior checks completed cleanly.
2 computersDiscovery, connection, substantive request/reply, repeat request, and clean queue.

Two self-contained executables were packaged. A second independent request reused the same tasks and sessions, and the acceptance ended with zero remaining test mail.

Result & next stage

A packaged, two-computer Windows prototype.

The system was tested, packaged, installed, and substantively live-verified between two Windows computers. Its current scope is exact-task coordination on a controlled local network; broader network and security validation is the next expansion stage.

Current live evidence covers discovery, connection, substantive request/reply, a repeated request, and a clean final queue across two Windows systems.
Next AI Systems Lab case studyPegasus Exchange