BREAKING CHANGE: GET /sse and POST /messages now answer 404. Every client
connects over Streamable HTTP at /mcp. The sse remote is removed from
server.json.
- Add unauthenticated GET /health ({status, version}); no session, no
rate-limit touch.
- Add any-origin CORS across the whole HTTP surface, incl. OPTIONS
preflight and error responses; mcp-session-id exposed to browsers.
- Unify every version string at 2.0.0; both McpServer identities now read
the version from package.json at build time instead of hardcoding it.
- Raise @modelcontextprotocol/sdk floor to ^1.30.0 and sync the local
install with the lockfile.
- Tests: retire the SSE-shaped auth-gating suite (its unique
process-liveness test is ported to streamable-http-gating), move the
shared harness and rate-limit suite to StreamableHTTPClientTransport,
and add http-surface.test.ts for CORS//health//sse-404 coverage.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
1.6 KiB
1.6 KiB
Changelog
2.0.0 - 2026-08-17
Breaking
- The legacy SSE transport is removed.
GET /sseandPOST /messagesnow answer 404. Every client connects over Streamable HTTP at/mcp:claude mcp add --transport http vector-mcp <url>/mcp. Thesseremote was removed from the MCP Registry manifest (server.json).
Added
GET /health: unauthenticated liveness probe, returns{"status":"ok","version":"<version>"}. It creates no session and bypasses the rate limiter.- Any-origin CORS on the whole HTTP surface, including preflight (
OPTIONS→ 204) and error responses, withmcp-session-idexposed to browser callers. Browser-based MCP clients can connect directly; authorization remains bearer-token only. Note for self-hosters: withMCP_AUTH_TOKENSunset, any web page can drive your instance from a visitor's browser, sharing that visitor's per-IP rate-limit bucket - setMCP_AUTH_TOKENSon any instance reachable by anyone but you.
Changed
- All package versions,
server.json, and both advertised MCP server identities are unified at 2.0.0; the identities now read the version frompackage.jsonat build time instead of hardcoding it. @modelcontextprotocol/sdkfloor raised to^1.30.0.
Known limits
- POST responses stream as SSE frames (the spec-default mode) with no keepalive bytes while a
tool call computes. Behind a proxy that cuts idle responses (Cloudflare's edge does at
~100 s), a single tool call that stays silent longer than that window is dropped by the
proxy. Long waits (e.g.
vector_await_transactionon a congested chain) should poll in shorter calls.