Roadmap
Phased build sequence
Section titled “Phased build sequence”
Phase 1: Foundation ✅
Section titled “Phase 1: Foundation ✅”Focus: Get Forge CLI to production quality. Ship publicly. Build initial community.
| Milestone | Status |
|---|---|
| Forge CLI v0.6 (production-ready) | ✅ Complete |
| Skills Starter v0.6 (feature-complete) | ✅ Complete |
| Website (crucibledx.dev) | ✅ Complete |
| npm publish (@crucibledx/forge-cli) | ✅ Complete |
| Public GitHub repos | ✅ Complete |
| CI/CD (GitHub Actions) | ✅ Complete |
| Homebrew tap + Scoop bucket | ✅ Complete |
Phase 2: Context Compiler (current)
Section titled “Phase 2: Context Compiler (current)”Focus: Evolve Forge from a file copier to a context compiler. Ship Ember MVP.
“Other tools install files. Forge compiles your AI environment.”
Each compiler feature ships as its own Forge version with room for patch releases:
| Version | Feature | Status |
|---|---|---|
| Forge v0.7 | Smithery/agentskills.io source adapter | 🟡 In progress |
| Forge v0.8 | Skill frontmatter schema + parser | ⬜ Designed |
| Forge v0.9 | Skill graph + dependency resolution | ⬜ Designed |
| Forge v0.10 | Project context detection + tree-shaking | ⬜ Designed |
| Forge v0.11 | Token budgeting + priority ordering | ⬜ Designed |
| Forge v0.12 | Per-agent emitters (compile targets) | ⬜ Designed |
| Forge v0.13 | Manifest generation (Ember bridge) | ⬜ Designed |
| Ember v0.1 | Manifest reader + CLI report | ⬜ Designed |
| Ember v0.2 | Multi-project discovery + aggregation | ⬜ Designed |
| Ember v0.3 | OTLP export + Grafana dashboard pack | ⬜ Designed |
Phase 3: Observability + Integrations
Section titled “Phase 3: Observability + Integrations”Focus: Grow Ember with external integrations. Drive adoption through existing communities.
| Version | Feature | Status |
|---|---|---|
| Ember v0.4 | Langfuse + Helicone integration | ⬜ Future |
| Ember v0.5 | Agent OTEL + file watchers (Claude Code, Cline) | ⬜ Future |
| Ember v0.6+ | Cursor/Windsurf/Copilot watchers | ⬜ Future |
| Forge v0.14+ | Code-splitting, incremental compilation | ⬜ Future |
Phase 4: Platform + Ecosystem
Section titled “Phase 4: Platform + Ecosystem”Focus: Extract shared infrastructure. Enable third-party plugins.
| Milestone | Status |
|---|---|
| @crucibledx/ember standalone package | ⬜ Future |
| @crucibledx/smelt extraction | ⬜ Future |
| @crucibledx/anvil SDK | ⬜ Future |
| Lathe MCP server | ⬜ Future |
| Foundry registry | ⬜ Future |
| Cast frontend CLI | ⬜ Future |
Architecture decisions
Section titled “Architecture decisions”| Decision | Choice | Rationale |
|---|---|---|
| Platform architecture | Path C (shared library) | Engineering leverage + community cohesion |
| Language | TypeScript/Bun | Fast iteration, single binary via compile |
| Plugin protocol | stdin/stdout JSON | Language-agnostic, easy to test |
| License | MIT | Maximum adoption |
| Compiler pipeline | In-process, pure domain | Parse → Analyze → Optimize → Emit, all pure functions |
| Ember data source | Forge compiler manifest | No instrumentation needed — build output as telemetry |
| Smelt extraction | Deferred | Don’t extract from one consumer |
| Go rewrite | Killed | TypeScript/Bun works, plugin protocol is language-agnostic |
Codebase metrics (Forge CLI v0.6.24)
Section titled “Codebase metrics (Forge CLI v0.6.24)”| Metric | Value |
|---|---|
| Production code | 6,369 LOC across 144 files |
| Unit tests | 6,700 LOC across 52 test files |
| E2E tests | 4,440 LOC across 25 test files |
| Total passing tests | 687 |
| Test-to-production ratio | ~1.75:1 |
| Releases | 66 versions (v0.1.0 → v0.6.24) |