v0.1.0 Release Notes¶
Release Date: 2026-05-24
Overview¶
MultiSpec v0.1.0 is the initial release providing multi-domain specification orchestration for humans and AI agents. This release establishes the foundation for bridging organizational intent (MRD, PRD, UXD) with executable specifications for AI coding agents.
Highlights¶
- Multi-domain specification orchestration - Separate specs for PM, UX, and Engineering with structured evaluation
- MCP server integration - Full integration with Claude Code and Kiro CLI via Model Context Protocol
- AI-assisted authoring workflow - Draft management with versioning, LLM evaluation, and iterative refinement
Features¶
Core Framework¶
- Type System: Core type definitions for spec types (MRD, PRD, UXD, Press, FAQ, Narrative, TRD, IRD), evaluation results, and project configuration
- Project Configuration:
multispec.yamlfor project-level settings and target configuration - Templates: Embedded spec templates using
go:embedfor all supported spec types
Draft Authoring¶
- Session Management: Start, resume, update, and finalize drafts with version tracking
- Evaluation History: Track evaluation attempts with scores, pass/fail status, and findings count
- Discard Support: Clean up abandoned drafts
Evaluation¶
- Categorical Rubrics: Pass/partial/fail scoring with weighted categories
- LLM Integration: Structured evaluation using configurable LLM providers
- Per-spec Rubrics: Custom rubrics for each spec type (MRD, PRD, UXD, Press, FAQ, Narrative, TRD, IRD)
Synthesis & Reconciliation¶
- GTM Synthesis: Generate press releases, FAQs, and narratives from source specs
- Technical Synthesis: Generate TRD and IRD from PRD and source specs
- Reconciliation: Combine approved specs into unified execution specification
Export¶
- Target Framework: Pluggable export system with capability detection
- SpecKit Adapter: Export to GitHub Spec-Kit format (spec.md, plan.md, tasks.md)
CLI Commands¶
| Command | Description |
|---|---|
init <project> |
Initialize a new project with standard directory structure |
lint [project] |
Validate directory structure and naming conventions |
status |
Show project status and readiness gates |
targets |
List available export targets |
eval [type] |
Evaluate specs using LLM judges |
synthesize <type> |
Generate GTM/technical specs from source docs |
reconcile |
Generate unified execution spec from approved specs |
export <target> |
Export to target execution system (SpecKit) |
MCP Server¶
Full MCP server implementation with the following tools:
| Tool | Description |
|---|---|
list_projects |
List all multispec projects |
get_project_status |
Get project readiness status |
start_draft |
Initialize a new draft |
update_draft |
Save draft content |
eval_draft |
Evaluate draft against rubric |
finalize_draft |
Promote draft to final spec |
get_draft |
Retrieve current draft |
discard_draft |
Delete a draft |
Skills¶
Authoring skills for Claude Code integration:
author-mrd- Guide MRD authoringauthor-prd- Guide PRD authoringauthor-uxd- Guide UXD authoring
Installation¶
Dependencies¶
- structured-evaluation v0.5.0 - Rubric and evaluation types
Platform Support¶
- Linux (amd64, arm64)
- macOS (amd64, arm64)
- Windows (amd64)
Known Limitations¶
- Export targets other than SpecKit are stubs
- Graphize integration not yet implemented
- No web UI for status visualization
What's Next¶
See ROADMAP.md for planned features including:
- Phase 2: Full evaluation engine with multi-judge support
- Phase 3: GTM and technical synthesis
- Phase 4: Reconciliation engine
- Phase 5: Additional target adapters
- Phase 7: Graphize integration for requirement graphs