feat: add --http-host flag for custom bind address

[r33drichards]

Summary

Adds --http-host CLI argument to allow binding the HTTP server to a specific network interface instead of the default 0.0.0.0 (all interfaces).

Changes

• Added --http-host CLI argument (default: 0.0.0.0)
• Updated start_streamable_http() function signature to accept host parameter
• Modified bind address construction to use format!("{}:{}", host, port)

Usage

# Bind to localhost only (secure, local access only)
server --http-port 3456 --http-host 127.0.0.1

# Bind to specific IP (e.g., dummy interface for subnet router)
server --http-port 3456 --http-host 172.31.255.1

# Default behavior (all interfaces)
server --http-port 3456

Use Cases

• Security: Bind to 127.0.0.1 to prevent external access
• Network segmentation: Bind to a specific interface IP
• Subnet router: Bind to a dummy interface for Tailscale subnet routing
• Multi-homed hosts: Choose which interface to listen on

Testing

Code compiles successfully (requires openssl-dev for full build). Changes are minimal and follow existing patterns in the codebase.

Backward Compatibility

✅ Fully backward compatible - defaults to 0.0.0.0 (existing behavior) when --http-host is not specified.

[github-actions]

MCP-V8 Load Test Benchmark Report

Comparison of single-node vs 3-node cluster at various request rates.

Results

Topology	Target Rate	Actual Iter/s	HTTP Req/s	Exec Avg (ms)	Exec p95 (ms)	Exec p99 (ms)	Success %	Dropped	Max VUs
cluster-stateful	100/s	62	1944.8	1491.82	4372.15	4541.28	100%	2017	100
cluster-stateful	200/s	63.4	3788	2953.77	5112.32	5167.58	67.7%	7876	200
cluster-stateless	1000/s	994.3	1719.2	54.54	176.6	740.22	24.3%	343	331
cluster-stateless	100/s	99.9	300.3	51.51	52.41	52.66	100%	3	13
cluster-stateless	200/s	199.5	601.9	52.73	52.84	60.01	100%	21	38
cluster-stateless	500/s	497.3	1270	45.66	53.88	111.59	71.1%	139	112
single-stateful	100/s	21.7	1902.4	4355.08	4802.42	4835.42	100%	4599	100
single-stateful	200/s	38.2	3760.6	4960.2	5188.36	5224.72	5.2%	9517	200
single-stateless	1000/s	199.4	1421.8	4845.28	10000.29	11906.79	95.8%	47704	1000
single-stateless	100/s	99.7	302.3	53.19	52.72	114.69	100%	11	21
single-stateless	200/s	199.1	600.3	53.74	53.97	96.66	100%	42	45
single-stateless	500/s	296	1502	1574.65	2864.33	3953.52	100%	11967	500

P95 Latency

Topology	Rate	P95 (ms)
cluster-stateful	100/s	4372.15	███████████████████████████
cluster-stateful	200/s	5112.32	████████████████████████████
cluster-stateless	100/s	52.41	█████████████
cluster-stateless	200/s	52.84	█████████████
cluster-stateless	500/s	53.88	█████████████
cluster-stateless	1000/s	176.6	█████████████████
single-stateful	100/s	4802.42	████████████████████████████
single-stateful	200/s	5188.36	████████████████████████████
single-stateless	100/s	52.72	█████████████
single-stateless	200/s	53.97	█████████████
single-stateless	500/s	2864.33	██████████████████████████
single-stateless	1000/s	10000.29	██████████████████████████████

Notes

• Target Rate: The configured constant-arrival-rate (requests/second k6 attempts)
• Actual Iter/s: Achieved iterations per second (each iteration = 1 POST /api/exec)
• HTTP Req/s: Total HTTP requests per second (1 per iteration)
• Dropped: Iterations k6 couldn't schedule because VUs were exhausted (indicates server saturation)
• Topology: single = 1 MCP-V8 node; cluster = 3 MCP-V8 nodes with Raft