fix: print why a service failed instead of naming a log file

[imran-siddique]

When cMCP Runtime fails to bind, the launchers wait the full 60 seconds and then say "See the *.log files in this demo folder." The reason is already in cmcp.log, and nobody looks there.

The case that costs the most time is a cmcp on PATH that is present but broken, which is easy to end up with alongside an editable dev install. I hit exactly this while verifying the claims on agentrust-io.com/demos: a minute of waiting, then a sentence pointing at a file.

Before

-- Starting cMCP Runtime (CMCP_DEV_MODE=1) on :8443 --
cMCP Runtime did not start listening on :8443 within 60s. See the *.log files in this demo folder.

After

-- Starting cMCP Runtime (CMCP_DEV_MODE=1) on :8443 --
cMCP Runtime exited with code 1 before listening on :8443.

Last 15 lines of cmcp.log:
    ...
      from agent_os.mcp_gateway import GovernancePolicy, MCPGateway
    ImportError: cannot import name 'GovernancePolicy' from 'agent_os.mcp_gateway'

Both branches are covered: the one where the process exits early, and the one where it never binds. Applied to the four demo launchers and the web console, which each carry their own copy of the wait helper by design, so each gets its own.

Verification

Reproduced a real failure and confirmed both branches print the tail. The success path is an untouched early return, so it is unchanged rather than re-verified.

Not fixed here

_find_cmcp() takes the first cmcp on PATH without checking that it runs, so a broken install still shadows a working .venv and the README's "it will use a local .venv if one exists" never engages. That needs a functional probe, and there is no cheap subcommand that exercises the same import start does: --version and validate-config both succeed against an install whose mcp.proxy import is broken. Worth a follow-up.

🤖 Generated with Claude Code

https://claude.ai/code/session_013EQx4N5BzTQbY8kvXUsdkY