Context
As discussed in Discussion #11002, we need a bounded release gate for v13 regarding the remote MCP transport. Instead of a broad epic covering all services, the triad swarm agreed on a single, focused release-blocking ticket to prove the substrate functionality over the container boundary.
The Problem
The current local-first architecture needs to be proven over remote Dockerized transport (SSE) to satisfy v13 requirements. We need a targeted validation of the remote MCP transport focusing on the read-heavy Memory Core and Knowledge Base services, avoiding the complexities of local-first filesystem modifications (GitHub Workflow) and live browser bridges (Neural Link).
The Architectural Reality
The proof will utilize the existing docker-compose.test.yml environment. The transport mechanism is SSE (/mcp endpoint). Identity assertion is handled out-of-band by cross-linking to the existing tests in OidcAuth.integration.spec.mjs (PR #10984), removing the need to re-verify auth token parsing in this specific proof.
The Fix
Implement an automated Playwright integration test that validates the Dockerized transport.
- Target the Knowledge Base (
neo-mjs-knowledge-base) and Memory Core (neo-mjs-memory-core) services.
- Dispatch at least one real tool call to each service via the remote transport using the host-side MCP SDK Streamable HTTP client.
- Confirm
Mcp-Session-Id header parsing and persistence inside the containerized services across multiple calls.
Acceptance Criteria
Out of Scope
neo-mjs-github-workflow service (local-first file modifications).neo-mjs-neural-link service (browser bridge complexity).- Re-implementation of Identity Token assertion (already covered by PR #10984).
Origin Session ID: d5ed6767-0292-46bf-9346-439f268048ec
Retrieval Hint: "Dockerized remote MCP transport proof v13 release gate"
Context As discussed in Discussion #11002, we need a bounded release gate for v13 regarding the remote MCP transport. Instead of a broad epic covering all services, the triad swarm agreed on a single, focused release-blocking ticket to prove the substrate functionality over the container boundary.
The Problem The current local-first architecture needs to be proven over remote Dockerized transport (SSE) to satisfy v13 requirements. We need a targeted validation of the remote MCP transport focusing on the read-heavy Memory Core and Knowledge Base services, avoiding the complexities of local-first filesystem modifications (GitHub Workflow) and live browser bridges (Neural Link).
The Architectural Reality The proof will utilize the existing
docker-compose.test.ymlenvironment. The transport mechanism is SSE (/mcpendpoint). Identity assertion is handled out-of-band by cross-linking to the existing tests inOidcAuth.integration.spec.mjs(PR #10984), removing the need to re-verify auth token parsing in this specific proof.The Fix Implement an automated Playwright integration test that validates the Dockerized transport.
neo-mjs-knowledge-base) and Memory Core (neo-mjs-memory-core) services.Mcp-Session-Idheader parsing and persistence inside the containerized services across multiple calls.Acceptance Criteria
test/playwright/integration/<name>.integration.spec.mjs.docker-compose.test.ymlsuccessfully boots KB and MC services./mcptool dispatch succeeds for at least one real KB tool beyond healthcheck./mcptool dispatch succeeds for at least one real MC tool beyond healthcheck.Mcp-Session-Idtest succeeds: init + same-session second-tool-call succeeds, proving session persistence.Out of Scope
neo-mjs-github-workflowservice (local-first file modifications).neo-mjs-neural-linkservice (browser bridge complexity).Origin Session ID: d5ed6767-0292-46bf-9346-439f268048ec Retrieval Hint: "Dockerized remote MCP transport proof v13 release gate"