MCP startup discovery ignores tools/list pagination

[Qiyuanqiii]

What happened?

MCP startup only discovers the first tools/list page. A server can connect successfully while tools on later pages are missing. An enabled_tools entry naming only a second-page tool produces an empty catalog.

connectStepMcpServer calls listTools once and ignores nextCursor.

Steps to reproduce

1. Serve a local HTTP MCP fixture: the first tools/list response contains tool_0 and nextCursor; the next page contains tool_1.
2. Connect through connectStepMcpServer.
3. Observe only one tools/list request and tool_0. The existing invokeRemoteMcpTool path correctly follows both pages against the same fixture.

The reproduction calls the production function directly without loading extensions.

Expected behavior

Collect every page before filtering and publishing the tools.

I have a small fix in my fork that reuses listAllMcpTools. Six new regression cases fail before the fix and pass afterwards; 33 focused tests and npm run check pass on Windows. Full Linux validation passed: build:offline, npm run check, and the complete isolated ./test.sh.

I'd like to submit the PR once approved with lgtm.

Version

Step Code 0.1.0, main @ cb5fc14; Node v24.14.0, Windows.

[github-actions]

This issue was auto-closed. All issues from new contributors are auto-closed by default.

Maintainers review auto-closed issues daily and reopen worthwhile ones. Issues that do not meet the quality bar in CONTRIBUTING.md will not be reopened or receive a reply.

If a maintainer replies lgtmi on one of your issues, your future issues will stay open. If a maintainer replies lgtm, your future issues and PRs will stay open. The command must be at the start of the reply (optionally after one or more @username mentions) or at the end.

See CONTRIBUTING.md.

[Qiyuanqiii]

AI-assisted contribution: Codex helped inspect the source, draft this report, implement the patch, and run the isolated reproduction and regression tests. The linked patch contains six new cases covering pagination, an empty intermediate page, filtering, cancellation, timeout, and publication only after all pages are collected. The full Linux validation result is available at the workflow link above.