← Research

first-party interoperability observations

Interoperability Evidence

A public ledger of what actually happened when Machine Realms and AIWebSignals were presented to external agent registries, scanners, and research-ingestion surfaces. Observed facts, inferred causes, remediation, and verification are kept separate.

This ledger does not treat listing, conformance, reachability, or a scanner result as proof of downstream agent use, endorsement, or representative ecosystem behavior.

mcp_registry · verified_success

Official MCP Registry

Observed: GitHub OIDC authentication and registry validation succeeded. The registry accepted io.github.Galanai/machinerealms version 1.0.0 with the remote streamable-http endpoint https://machinerealms.com/mcp.

Expected: Registry metadata validates, repository ownership is authorized, and the remote MCP server becomes discoverable.

Remediation: Shortened the registry description and matched the exact case-sensitive GitHub OIDC namespace.

Verification: Registry API returned one active result with isLatest=true.

Evidence and limitations

Registry acceptance proves publication and discoverability metadata, not downstream agent use.

a2a · verified_success

A2A Registry

Observed: The registry accepted MachineRealms and reported Fetched, Conformant, Reachable, 100% reachability, with JSON-RPC and HTTP+JSON interfaces declared.

Expected: The public Agent Card is fetched, validated, and the declared endpoint is reachable.

Remediation: Added a JSON-RPC A2A endpoint at /a2a while preserving /a2a/v1 as the HTTP+JSON alternate.

Verification: Registry listing reported conformant and reachable after the JSON-RPC binding was deployed.

Evidence and limitations

Registry conformance and reachability do not prove task completion or repeated agent use.

a2a · reproduced_and_remediated

A2A Registry

Observed: The registration flow rejected the Agent Card when HTTP+JSON was the only declared binding and required a JSONRPC-compatible interface.

Expected: A standards-oriented A2A registry accepts the advertised Agent Card.

Remediation: Added JSON-RPC compatibility and re-registered.

Verification: Subsequent registry registration succeeded and became publicly listed.

Evidence and limitations
  • operator_observed_registry_validation: A2A Registry public ingestion run, 2026-09-19

This is evidence about one registry implementation and should not be generalized to every A2A consumer.

mcp · remediation_in_progress

Smithery

Observed: Smithery discovered server metadata, then reported "Initialization failed with status 400" and could not automatically scan the server.

Expected: Smithery automatically scans the public unauthenticated MCP endpoint and enumerates capabilities.

Remediation: Added backwards-compatible initialize/tools support, read-only tool wrappers, and a static server-card endpoint.

Verification: Pending a fresh Smithery publication scan after the compatibility deployment.

Evidence and limitations
  • operator_observed_release_log: Smithery release log at 10:14:28 local time on 2026-09-19
  • public_endpoint

The causal explanation is inferred from protocol behavior and scanner output; Smithery has not independently confirmed the cause.

mcp · observed_unresolved_external_behavior

Smithery

Observed: Smithery displayed an "Authorization Required" modal after scan failure even though MachineRealms discovery metadata explicitly declared authentication.required=false and the endpoint does not require credentials.

Expected: A public no-auth MCP endpoint is represented as unauthenticated.

Remediation: Do not authorize a nonexistent credential flow; instead expose scanner-compatible protocol behavior and a complete static server card.

Verification: No credential requirement exists in the MachineRealms MCP contract. Smithery rescan result remains pending.

Evidence and limitations

This describes observed Smithery UI behavior, not a claim about Smithery intent or all scanner states.

mcp · implemented_pending_external_verification

Smithery

Observed: Smithery explicitly instructed the publisher to advertise /.well-known/mcp/server-card.json to skip automatic scanning.

Expected: A scanner that cannot initialize the endpoint can still discover tools from a static manifest.

Remediation: Published a static card with serverInfo, no-auth declaration, top-level tools/resources/prompts, transport metadata, protocol versions, and read-only tool annotations.

Verification: First-party endpoint is live; Smithery publication acceptance remains pending.

Evidence and limitations

The exact static-card schema is an interoperability convention and may evolve.

mcp · published_external_index_stale

Smithery

Observed: The public Smithery listing resolves for keyanshokraie/machinerealms but still displays No description, No capabilities found, and No deployments found while the live MachineRealms server card exposes a no-auth streamable-http endpoint with a complete read-only tool and resource catalog.

Expected: A published Smithery listing reflects the tools and resources available from the live public MCP server card.

Remediation: No additional credential or authority change is warranted. The compatibility surface and static card are live; Smithery exposes no manual rescan action, so the remaining state is external metadata refresh/indexing.

Verification: Authenticated browser inspection confirmed the listing is published and the toolbox connection exists; direct server-card inspection confirmed the current tool/resource metadata.

Evidence and limitations

This observation does not establish how frequently Smithery refreshes listings or whether its background index will update without another publisher-side action.

distribution_directory · human_authorization_required

Glama

Observed: The Add Server flow requires an authenticated user account before submission. No anonymous server-add path was observed.

Expected: A free MCP directory may allow server submission after publisher authentication.

Remediation: Publisher authentication is required; no protocol or MachineRealms runtime change is indicated.

Verification: Observed sign-in boundary before server submission.

Evidence and limitations

Account requirements and UI flows can change.

agent_research_ingestion · verified_failure_current

AIWebSignals first-party research collector

Observed: GET requests to /api/agent-experience/observations and /api/agent-reflective/claims returned HTTP 503 during verification.

Expected: Published agent-experience and reflective research contracts can accept controlled observation submissions.

Remediation: Provision the collector secret and verify controlled ingestion without weakening the trust boundary.

Verification: Rechecked on 2026-09-19; both endpoints still returned 503.

Evidence and limitations

This proves current intake unavailability, not absence of machine traffic to other AIWebSignals interfaces.