v0.1.64

🚀 Release Highlights — v0.1.64 (2026-03-16)

🔌 Gemini CLI Backend

• Gemini CLI as a native backend: Google's Gemini CLI with subprocess-based streaming
• Session persistence: Multi-turn conversations via CLI session IDs
• MCP tools: Wired through .gemini/settings.json with native hook adapter for tool execution
• Docker support: Containerized execution via gemini_cli_docker.yaml config

🔍 Execution Trace Analyzer

• New subagent type: Mechanistic analysis of agent execution traces to extract durable learnings
• 7-dimension evaluation: Error learning, effort allocation, approach effectiveness, tool strategy, reasoning patterns, context health, verification completeness
• Output: process_report.md (narrative) and process_verdict.json (structured scores)

⚡ WebSocket Streaming

• Persistent WebSocket transport: wss:// connection to OpenAI Response API for real-time event streaming
• Auto-reconnection: Configurable retry logic with exponential backoff
• YAML config: Enable with websocket_mode: true on OpenAI backend

🐳 Copilot Docker Mode

• Containerized tool execution: command_line_execution_mode: "docker" for Copilot backend
• Configuration: Docker sudo support, network mode selection (bridge/host)

✅ Fixes

• Response API duplicates: Prevent duplicate item errors in recursive tool loops (#1000)

---

📖 Getting Started

• Quick Start Guide
• Try It:

  pip install massgen==0.1.64
  # Try the Gemini CLI backend
  uv run massgen --config @examples/providers/gemini/gemini_cli_local "Explain quantum computing"

What's Changed

• Trace eval sub by @ncrispino in #1002
• feat: add websocket_mode for responses api by @praneeth999 in #990
• fix: Response API duplicate item errors and Windows MCP server crashes by @db-ol in #1000
• feat: Gemini cli clean by @ncrispino in #999
• docs: docs for v0.1.64 by @Henry-811 in #1003
• feat: v0.1.64 by @Henry-811 in #998

Full Changelog: v0.1.63...v0.1.64

v0.1.63

🚀 Release Highlights — v0.1.63 (2026-03-13)

🎯 Ensemble Pattern

• Ensemble defaults for subagents: disable_injection and defer_voting_until_all_answered now default to true, so subagents work independently before voting for more diverse, higher-quality results
• Automatic ensemble orchestration: Defaults apply when spawning subagent orchestrators without explicit override

🔄 Round Evaluator Improvements

• Transformation pressure: Evaluator pushes agents toward meaningful structural changes rather than surface-level edits
• Success contracts: Explicit quality gates agents must satisfy before the round evaluator allows convergence
• Verification replay: Evaluation consistency across rounds via replayed verification context

⚡ Lighter Refinement

• Reduced subagent overhead: Lighter refinement prompts for subagent workflows cut token usage and latency
• Killed agent handling: Graceful management of agents that time out or fail mid-round

✅ Fixes

• Timeout fallback: More robust coordination when agents hit timeout boundaries

---

📖 Getting Started

• Quick Start Guide
• Try It:

  pip install massgen==0.1.63
  # Try the round evaluator with ensemble defaults
  uv run massgen --config @examples/features/round_evaluator_example.yaml "Create a polished landing page for an AI product"

What's Changed

• feat: Better subagent contracts; lighter refinement for subagents too by @ncrispino in #996
• docs: docs for v0.1.63 by @Henry-811 in #997
• feat: v0.1.63 by @Henry-811 in #995

Full Changelog: v0.1.62...v0.1.63

v0.1.62

🚀 Release Highlights — v0.1.62 (2026-03-11)

🧩 MassGen Skill

• Multi-agent collaboration as a skill: Install with npx skills add massgen/skills --all and use MassGen directly from Claude Code, Cursor, Copilot, and 40+ other AI agents
• Four modes: General (any task), Evaluate (critique existing work), Plan (structured project plans), Spec (requirements specifications)
• Auto-distributed: Skill automatically syncs to a dedicated repository for easy installation

👁️ Session Viewer

• Watch automation runs in real-time: New massgen viewer command opens a TUI to observe running or completed sessions
• Session picker: --pick flag for browsing and selecting specific sessions, --web for browser-based viewing

⚡ Backend & Quickstart Improvements

• Claude Code backend: Background task execution and native MCP support via the SDK
• Codex backend: Native filesystem access and MCP tool integration
• Copilot backend: Runtime model discovery with automatic capability detection
• Headless quickstart: Non-interactive setup via --quickstart --headless for CI/CD pipelines
• Web quickstart: Browser-based setup via --web-quickstart

✅ Fixes

• Evaluation criteria: Removed should/could criteria that caused agents to produce overly similar outputs
• Planning prompts: Improved planning prompts with configurable thoroughness levels

---

📖 Getting Started

• Quick Start Guide
• Try It:

  # Install the MassGen Skill for your AI agent
  npx skills add massgen/skills --all
  # Then use MassGen from Claude Code, Cursor, Copilot, etc.
  
  # Or install MassGen directly and try the Session Viewer
  pip install massgen==0.1.62
  uv run massgen viewer --pick

What's Changed

• feat: MassGen skill by @ncrispino in #992
• docs: docs for v0.1.62 by @Henry-811 in #993
• feat: v0.1.62 by @Henry-811 in #991

Full Changelog: v0.1.61...v0.1.62

v0.1.61

🚀 Release Highlights — v0.1.61 (2026-03-09)

🔄 Round Evaluator Paradigm

• Automatic post-answer evaluation: New round_evaluator subagent type that automatically spawns evaluator subagents after each new answer, feeding detailed feedback into the next round
• Configurable evaluation flow: Control whether evaluation runs before or after checklist grading, whether to skip synthesis, and whether evaluators refine their feedback
• Example config: New round_evaluator_example.yaml — one agent builds while three agents evaluate in parallel

📝 Evaluation & Prompt Improvements

• Task plan injection: Evaluation prompts now include the current task plan for context-aware quality assessment
• Clearer evaluation prompts: Rewritten round evaluation prompts for more actionable, focused feedback

✅ Fixes

• Session resumption: Fixed crash when resuming from an already-resumed log
• Timeout fallback: When coordination times out, the latest answer is used directly without an extra final presentation step
• Subagent compatibility: Improved SUBAGENT.md template for broader subagent type support
• Codex backend: Added Codex backend support for new orchestrator features

---

📖 Getting Started

• Quick Start Guide
• Try It:

  # Install or upgrade to v0.1.61
  pip install --upgrade massgen
  
  # One agent builds, 3 agents evaluate — round evaluator in action
  uv run massgen --config @examples/features/round_evaluator_example.yaml "Create a website about a fictional AI product that is visually stunning and has at least one unique interactive element"

What's Changed

• feat: Add round evaluator paradigm by @ncrispino in #986
• docs: docs for v0.1.61 by @Henry-811 in #987
• feat: v0.1.61 by @Henry-811 in #985

Full Changelog: v0.1.60...v0.1.61

v0.1.60

🚀 Release Highlights — v0.1.60 (2026-03-06)

🛠️ Multimodal Tool Improvements

• Rewritten read_media: Clearer tool schema, better error handling, and improved naming for more reliable media understanding
• Media Call Ledger: Automatic tracking of all read_media and generate_media calls for observability

🤖 Subagent Enhancements

• Backend Inheritance: New inherit_spawning_agent_backend option lets subagents automatically use the same backend as the agent that spawned them
• Final Answer Strategy: New final_answer_strategy option controls how subagent orchestrators select the final answer (reuse winner, have winner present, or synthesize)
• Per-Agent Subagent Configs: Each agent can now define its own subagent_agents override for fine-grained subagent control

🧠 GPT-5.4 Support

• New default OpenAI flagship: GPT-5.4 added to the model registry, ready to use across all coordination modes

🔄 Decomposition + Checklist Cooperation

• Unified Quality Workflow: Decomposition mode now cooperates with the checklist workflow, enabling quality-gated subtask iteration
• Faster Verification Rounds: Improved prompts for verification replay, reducing verification round time

✅ Fixes

• Checklist & Prompt Injection: More reliable checklist behavior with improved proposal injection and system prompt refocused on entire output quality
• Codex Pricing Accuracy: Fixed prompt caching calculation for correct cost tracking
• Task Plan Refresh: Fixed plan refresh during quality rounds
• Skill Prefix Handling: Fixed edge cases in skill prefix resolution

---

📖 Getting Started

• Quick Start Guide
• Try It:

  # Install or upgrade to v0.1.60
  pip install --upgrade massgen
  
  # Choose backend 'openai' with model 'gpt-5.4' in the setup wizard to start using GPT-5.4
  uv run massgen --quickstart

What's Changed

• feat: Improve verification time by @ncrispino in #978
• docs: docs for v0.1.60 by @Henry-811 in #979
• feat: v0.1.60 by @Henry-811 in #974

Full Changelog: v0.1.59...v0.1.60

v0.1.59

🚀 Release Highlights — v0.1.59 (2026-03-04)

🔄 Smarter Quality Rounds

• Verification Replay Memories: Agents save replayable verification steps (commands, scripts, artifacts) to verification_latest.md, auto-injected into future rounds so the next agent can replay the exact verification pipeline
• Plan-Tracked Improvements: Improvements from each round are auto-added to the task plan, so agents build on prior progress instead of starting fresh
• Enhanced Plan Review: More thorough quality evaluation during plan review phases

✅ Checklist & Evaluation Fixes

• More Accurate Evaluations: Improved evaluation generation config for higher-quality assessments
• Consistent Checklist Behavior: Fixed checklist handling across rounds
• Gemini MCP Compatibility: Tool name normalization for Gemini backends using MCP tools

🔧 Infrastructure

• Subagent Enhancements: Improved coordination, task delegation, and Docker skill write access
• Video Generation Fixes: Cleaner error handling (no silent fallback to animated), restored impact metrics
• Bug Fixes: Answer anonymization fix, quickstart updates

---

📖 Getting Started

• Quick Start Guide
• Try It:

  # Install or upgrade to v0.1.59
  pip install --upgrade massgen
  
  # Try checklist-gated quality rounds with verification replay
  uv run massgen --config @examples/features/subagent_checklist.yaml \
    "Create a website for an AI company selling a creative sci-fi style product. Ensure polished visuals and cool interactive elements"

What's Changed

• feat: Improve quality rounds by @ncrispino in #969
• docs: docs for v0.1.59 by @Henry-811 in #970
• feat: v0.1.59 by @Henry-811 in #968

Full Changelog: v0.1.58...v0.1.59

v0.1.58

🚀 Release Highlights — v0.1.58 (2026-03-02)

🖼️ Comprehensive Multimodal Revamp

• New Media Providers: ElevenLabs (TTS/STT), Nano Banana 2 (default image gen), and Grok Imagine (image/video) join existing providers
• Media Generation Skills: Reusable skills for image, video, and audio generation workflows
• Multi-Turn Image Editing: Iterative image editing for supported providers — agents can refine images across rounds

🟢 Nvidia NIM Backend

• NVIDIA Inference Microservices: First-class provider for NVIDIA-hosted models via NIM API

🔍 Quality Rethinking Subagent

• Per-Element Craft Improvements: New quality_rethinking subagent type that targets specific elements for refinement
• Improve/Preserve Checklists: Checklists now explicitly separate what to improve vs. what to preserve

🖥️ CLI Mode Flags

• New Flags: --quick, --single-agent, --coordination-mode, --personas flags mirror TUI toggles
• Plan Mode from CLI: Start plan mode directly from the command line

🔧 Infrastructure

• Logging Refactor: Fixed concurrent logging for parallel multi-agent execution — each agent gets isolated log context
• Subagent Hardening: Better error handling for malformed inputs and repeated tool calls
• Evaluation Criteria Defaults: Sensible defaults when criteria are not explicitly specified

---

📖 Getting Started

• Quick Start Guide
• Try It:

  # Install or upgrade to v0.1.58
  pip install --upgrade massgen
  
  # Try checklist-driven refinement with quality rethinking
  uv run massgen --config @examples/features/subagent_checklist.yaml \
    "Create a website for an AI company selling a creative sci-fi style product. Ensure polished visuals and cool interactive elements"

What's Changed

• feat(backend): add Nvidia NIM as a first-class backend provider by @AbhimanyuAryan in #962
• feat: Improve coordination with improvements; improve and expand multimedia generation by @ncrispino in #964
• docs: docs for v0.1.58 by @Henry-811 in #965
• feat: v0.1.58 by @Henry-811 in #957

Full Changelog: v0.1.57...v0.1.58

v0.1.57

🚀 Release Highlights — v0.1.57 (2026-02-27)

🔗 Subagent Delegation Protocol

• Container-to-Host Spawning: File-based delegation via SubagentLaunchWatcher with atomic JSON request/response exchange and workspace path validation
• Auto-Mounted Parent Workspace: Subagents get parent workspace (read-only) by default

🏗️ Builder Subagent

• Fresh-Context Artifact Generation: New subagent type for transformative redesigns and complex multi-file rewrites with prescriptive specs

📊 Smarter Convergence

• Substantiveness Tracking: Planned changes classified as transformative/structural/incremental — triggers builder or novelty subagent accordingly
• Diagnostic Report Gating: Optional quality gate — agents must submit a structured diagnostic report before checklist passes
• Per-Agent Checklist Scoring: Evaluate multiple agents separately with automatic format detection

🔧 Infrastructure

• Claude Code Reasoning: Unified reasoning config (type, effort, budget_tokens) replacing deprecated max_thinking_tokens
• Bug Fixes: Fixed Codex subagent spawning, subagent synchronization/timeout handling, temp workspace directories

---

📖 Getting Started

• Quick Start Guide
• Try It:

  # Install or upgrade to v0.1.57
  pip install --upgrade massgen
  
  # Try builder subagent with checklist-driven refinement
  uv run massgen --config @examples/features/subagent_checklist.yaml \
    "Create a website for an AI company selling a creative sci-fi style product. Ensure polished visuals and cool interactive elements"

What's Changed

• feat: Improve subagent calling, eval criteria by @ncrispino in #955
• docs: docs for v0.1.57 by @Henry-811 in #956
• feat: v0.1.57 by @Henry-811 in #947

Full Changelog: v0.1.56...v0.1.57

v0.1.56

🚀 Release Highlights — v0.1.56 (2026-02-25)

📋 Spec Plan Mode

• Formal Requirements Before Execution: plan_mode="spec" — spec creation, approval modal, and execution pipeline with changedoc integration. TUI spec mode (Shift+Tab twice to enter)

🎯 ask_others Targeted Messaging

• Focused Agent-to-Agent Communication: target_agents parameter for directed queries instead of broadcast, with per-target validation and response counting

🔧 Infrastructure

• Critic Subagent: New subagent type for honest quality assessment — detects genuine vs incremental improvement, quality ceiling. Enhanced novelty subagent guidance for growth-oriented refinement
• read_media Conversation Continuity: Follow-up image conversations via continue_from conversation_id
• Codex OAuth Login Fix: Codex backend always available in WebUI regardless of OPENAI_API_KEY
• Docker Configuration Mounting: Claude and Codex configuration mounting for Docker containers

---

📖 Getting Started

• Quick Start Guide
• Try It:

  # Install or upgrade to v0.1.56
  pip install --upgrade massgen
  
  # Launch MassGen, then press Shift+Tab twice to enter 'spec' mode
  uv run massgen

What's Changed

• fix: adding codex for OAuth login by @MuL1ian in #937
• feat: Add spec mode by @ncrispino in #945
• docs: docs for v0.1.56 by @Henry-811 in #946
• feat: v0.1.56 by @Henry-811 in #944

Full Changelog: v0.1.55...v0.1.56

v0.1.55

🚀 Release Highlights — v0.1.55 (2026-02-23)

🧩 Specialized Subagent Types

• Discovery-Based Roles: Specialized subagent roles via SUBAGENT.md frontmatter — evaluator (programmatic verification), explorer (investigation), researcher (deep analysis), novelty (breaks refinement plateaus)
• TUI Visualization: Subagent roles displayed in agent status indicators

📊 Dynamic Evaluation Criteria

• Task-Specific Quality Gates: GEPA-inspired evaluation criteria generation replacing static checklist items, with domain-specific presets (persona, decomposition, evaluation, prompt, analysis)
• Core/Stretch Convergence: Items categorized as core or stretch — convergence off-ramp triggers when all core items pass. Score scale 0-10. Config: evaluation_criteria_generator

🔧 Infrastructure

• Native Backend Image Routing: understand_image routes to agent's own backend (Claude, Gemini, Grok, Claude Code, Codex) with OpenAI fallback
• Configurable Video Frame Extraction: Scene-based (PySceneDetect) or uniform modes with max_frames cost guardrail (default 30, max 60). Config: multimodal_config.video
• Remotion Skill: Video generation/editing skill available in quickstart wizard
• Unified Pre-Collaboration: Persona generation, decomposition, and eval criteria generation unified as composable primitives

---

📖 Getting Started

• Quick Start Guide
• Try It:

  # Install or upgrade to v0.1.55
  pip install --upgrade massgen
  
  # Multi-agent coordination with specialized subagents
  uv run massgen --config massgen/configs/features/background_subagent_example.yaml --cwd-context ro "Use an explorer subagent to analyze this repo"

What's Changed

• feat: Add subagent roles by @ncrispino in #938
• docs: docs for v0.1.55 by @Henry-811 in #939
• feat: v0.1.55 by @Henry-811 in #935

Full Changelog: v0.1.54...v0.1.55