- CLI subcommands:
doctor,manifest,run,init - Production features default-on: worker-pool, metrics, config-watch
- Documentation: SECURITY.md, CONTRIBUTING.md, examples/
- Production configuration template
- SHA256 installer verification, doctor enhancements, structure-aware truncation
- Concurrent server startup via JoinSet + Semaphore
TransportDeaderror variant distinguishing permanent transport failures from transient errorsReconnectingClientdecorator with exponential backoff auto-reconnection- Per-server
reconnectandmax_reconnect_backoff_secsconfiguration - rmcp 1.2.0 migration (
#[non_exhaustive]structs, builder API) - deno_core 0.398, V8 147.2, oxc 0.126, sha2 0.11 (v0.5.1)
Notify-coordinated reconnection (v0.5.1)
- Prometheus metrics endpoint (HTTP)
- OpenTelemetry tracing integration
- Structured audit log output (JSON)
- Health check endpoint
- Dashboard templates (Grafana)
- Server discovery (auto-detect MCP servers on PATH)
- Configuration profiles (dev/staging/production)
- Plugin system for custom validators
- Stable public API (SemVer guarantee)
- Long-term support commitment
- Comprehensive migration guides
- Performance benchmarks and optimization
The following are explicitly out of scope:
- VS Code / IDE extension — Forgemax is a protocol-level gateway, not an editor plugin
- GUI / TUI — CLI-first design; use
forgemax doctor --jsonand pipe to your preferred UI - Telemetry / analytics — No data leaves your machine unless you configure a downstream server
- Built-in lodash / utility libraries in sandbox — Keep the sandbox minimal; LLMs can write vanilla JS
- Plugin system for sandbox extensions — Security boundary must remain simple and auditable