[#2] Stdio server spawn-test #2

New Issue

2026-07-02T00:49:12-04:00

the_og commented

2026-07-02 00:49:12 -04:00

Problem

check_dependency() only verifies whether command resolves on PATH — it does not verify the server actually starts. The single most common troubleshooting step people do by hand is "run the server manually to see if it starts without errors" (MCP official debugging docs). BCC stops one step short of that.

Proposed solution

Add a Test launch function in bcc_core.py that:

Spawns the process using subprocess with the server's command, args, and env.
Waits up to a short timeout (e.g. 3 s) for the process to crash or produce output.
Captures stderr (MCP servers must not write non-JSON-RPC content to stdout, so stderr is the right debug channel).
Returns a result dict: {outcome: "ok"|"crashed"|"timeout", stderr: str, returncode: int|None}.

The GUI (bcc.py) adds a Test launch button next to each stdio server row; clicking it runs the above in a QThread (same pattern as ConnTester) and shows the result inline (green/red) with a Copy output action.

Acceptance criteria

Core function is in bcc_core.py with no GUI imports; has unit tests in tests/test_core.py.
Spawning a valid server (e.g. python -c "import time; time.sleep(10)") → timeout.
Spawning an immediately-crashing command → crashed with stderr.
Spawning command-that-does-not-exist → handled gracefully (not a Python exception bubble).
Button appears per stdio server in the GUI; result shown inline; stderr tail available via Copy diagnostics.

Relevant code

bcc_core.py::check_dependency (~line 916) — PATH check to reuse/extend
bcc_core.py::runner_hint (~line 885)
bcc.py::ConnTester — pattern to follow for the QThread wrapper

## Problem `check_dependency()` only verifies whether `command` resolves on PATH — it does not verify the server actually starts. The single most common troubleshooting step people do by hand is "run the server manually to see if it starts without errors" (MCP official debugging docs). BCC stops one step short of that. ## Proposed solution Add a **Test launch** function in `bcc_core.py` that: 1. Spawns the process using `subprocess` with the server's `command`, `args`, and `env`. 2. Waits up to a short timeout (e.g. 3 s) for the process to crash or produce output. 3. Captures **stderr** (MCP servers must not write non-JSON-RPC content to stdout, so stderr is the right debug channel). 4. Returns a result dict: `{outcome: "ok"|"crashed"|"timeout", stderr: str, returncode: int|None}`. The GUI (`bcc.py`) adds a **Test launch** button next to each stdio server row; clicking it runs the above in a `QThread` (same pattern as `ConnTester`) and shows the result inline (green/red) with a **Copy output** action. ## Acceptance criteria - Core function is in `bcc_core.py` with no GUI imports; has unit tests in `tests/test_core.py`. - Spawning a valid server (e.g. `python -c "import time; time.sleep(10)"`) → `timeout`. - Spawning an immediately-crashing command → `crashed` with stderr. - Spawning `command-that-does-not-exist` → handled gracefully (not a Python exception bubble). - Button appears per stdio server in the GUI; result shown inline; `stderr` tail available via **Copy diagnostics**. ## Relevant code - `bcc_core.py::check_dependency` (~line 916) — PATH check to reuse/extend - `bcc_core.py::runner_hint` (~line 885) - `bcc.py::ConnTester` — pattern to follow for the QThread wrapper

the_og added the P0 label 2026-07-02 00:49:12 -04:00

the_og referenced this issue

2026-07-02 00:50:22 -04:00

[#6] In-app MCP server log viewer #6

the_og referenced this issue from a commit

2026-07-02 00:57:51 -04:00

feat: stdio server spawn-test (issue #2)

the_og referenced a pull request that will close this issue

2026-07-02 00:58:09 -04:00

feat: stdio server spawn-test (#2) #11

the_og referenced this issue

2026-07-02 01:03:24 -04:00