feat(stdio): explain initialize-time EOF startup failure (YTMCP-23)

Claude-Run commented

2026-06-25 11:37:48 +02:00

Member

When stdio mode is launched with no MCP client owning stdin (e.g. as a systemd service with stdin from /dev/null, or run interactively from a tty), rmcp's serve() returns the opaque connection closed: initialize request and the process exits non-zero, which a service manager then restart-loops with no signal about the real cause.

Add a pre-flight WARN when stdin is a terminal, and map the initialize-time EOF (rmcp's ServerInitializeError::ConnectionClosed with context "initialize request") to an actionable ERROR before propagating. The diagnostic names the cause (no MCP client on stdin), the fix (let the client launch it, or set YT_MCP_TRANSPORT=http for a service, cross-referencing docs/deployment.md), and notes the token is a red herring because the failure is at the transport handshake before any YouTrack call. The process still exits non-zero so the misconfiguration is not masked. Covered by unit tests for the predicate and diagnostic content.

Fixes #YTMCP-23

When stdio mode is launched with no MCP client owning stdin (e.g. as a systemd service with stdin from /dev/null, or run interactively from a tty), rmcp's serve() returns the opaque `connection closed: initialize request` and the process exits non-zero, which a service manager then restart-loops with no signal about the real cause. Add a pre-flight WARN when stdin is a terminal, and map the initialize-time EOF (rmcp's ServerInitializeError::ConnectionClosed with context "initialize request") to an actionable ERROR before propagating. The diagnostic names the cause (no MCP client on stdin), the fix (let the client launch it, or set YT_MCP_TRANSPORT=http for a service, cross-referencing docs/deployment.md), and notes the token is a red herring because the failure is at the transport handshake before any YouTrack call. The process still exits non-zero so the misconfiguration is not masked. Covered by unit tests for the predicate and diagnostic content. Fixes #YTMCP-23

Claude-Run added 1 commit

2026-06-25 11:37:48 +02:00

feat(stdio): explain initialize-time EOF startup failure (YTMCP-23)

Check / fmt + clippy + build + tests (pull_request) Successful in 23s

Details

Create release / Create release from merged PR (pull_request) Has been skipped

Details

93bae7fe5c

When stdio mode is launched with no MCP client owning stdin (e.g. as a systemd service with stdin from /dev/null, or run interactively from a tty), rmcp's serve() returns the opaque `connection closed: initialize request` and the process exits non-zero, which a service manager then restart-loops with no signal about the real cause.

Add a pre-flight WARN when stdin is a terminal, and map the initialize-time EOF (rmcp's ServerInitializeError::ConnectionClosed with context "initialize request") to an actionable ERROR before propagating. The diagnostic names the cause (no MCP client on stdin), the fix (let the client launch it, or set YT_MCP_TRANSPORT=http for a service, cross-referencing docs/deployment.md), and notes the token is a red herring because the failure is at the transport handshake before any YouTrack call. The process still exits non-zero so the misconfiguration is not masked. Covered by unit tests for the predicate and diagnostic content.

#YTMCP-23

David merged commit 8eac991c97 into main

2026-06-25 11:46:17 +02:00

David referenced this pull request from a commit

2026-06-25 11:46:18 +02:00

Merge pull request 'feat(stdio): explain initialize-time EOF startup failure (YTMCP-23)' (#30) from feat/ytmcp-23-stdio-startup-diagnostics into main

David deleted branch feat/ytmcp-23-stdio-startup-diagnostics

2026-06-25 11:46:18 +02:00

Rows
Columns

feat(stdio): explain initialize-time EOF startup failure (YTMCP-23) #30