RoadmapWhat ships now.
What we're willing to commit to.
No vague "coming soon" markers. Shipped items have a date. Planned items name the release they land in. Future items are explicit about being directional, not promises.
v1.0.0
Shipped2026-04-21First public release- 88 specialist agents across 7 executive offices
- 34 domain skills, each citing a behavioral framework
- 38 slash commands / 7 MCP servers / 47 MCP tools
- 64 behavioral frameworks cataloged in FRAMEWORKS_MASTER
- 14 compliance frameworks enforced at the MCP layer
- 7 public ADRs published as governance contracts
- 6 L1 commercial packs + 4 L2 bundles + 4 L3 vertical overlays
- 7 community packs auto-install on first session (150 VoltAgent specialists + Superpowers + Planning With Files + Claude Skills Library + UI/UX Pro Max + Higgsfield + 59 brand design templates)
- Cloudflare Worker license issuance (RS256 JWT + machine fingerprint) live
- Stripe-primary payment processor per ADR-011
- Three-layer .brand-foundation voice system
- Proactive Watch running 10 drift classes every session start
- Business Source License 1.1: converts to Apache 2.0 on 2030-04-21
v1.0.1
Shipped2026-04-27MCP path bugfix + count alignment + Class 11- BUG-006 fix: all 7 plugin:maxim:mxm-* MCPs now connect (relative path → ${CLAUDE_PLUGIN_ROOT} substitution)
- Agent count: 88 → 90 (cost-analyst CINO + sre-analyst COO + skill-synthesizer drift fix)
- Class 11 surface-claims-drift codified: Proactive Watch now scans 11 classes
- bootstrap/sync-counts.{sh,ps1} tool: propagates capability counts across 30+ surfaces idempotently
v1.1.0
Shipped2026-04-27License middleware foundation- Cloudflare Worker /issue + /validate endpoints LIVE at maxim-license-api.isystematic.workers.dev
- All 7 MCP servers gated via mcp/_shared/license-gate.mjs (RS256 JWT + machine fingerprint)
- v1.1.0.1: 90-day Pro Trial auto-activates on first install; falls back to Starter Day 91+
- v1.1.0.2: BUG-007 fix: plugin-scoped MCP-deps install + sentinel
- v1.1.0.3: Single-restart upgrade via spawn-with-deps wrapper
- v1.1.0.4: /mxm-help content drift fix
v1.1.1
Shipped2026-04-28Self-update + public install docs- /mxm-self-update slash command: pulls latest from marketplace, syncs into install cache, preserves node_modules
- documents/INSTALL.md: public install/upgrade/uninstall guide covering base + 14 packs
- Tester upgrade flow: ~5 sec + 1 restart (was ~2 min + 2 restarts)
v1.1.2
PlannedQ3 2026Maxim Overlay Engine + 7 compliance frameworks- Maxim Overlay Engine (MOE) per ADR-012: apply Maxim governance to every Claude Code plugin output via 4-hook architecture (SessionStart / UserPromptSubmit / PreToolUse / PostToolUse)
- 7 new compliance frameworks: EU AI Act, ISO 42001, SOX, CIS Controls, DORA, NIST SP 800-53, LGPD
- Tier-gated overlay behavior (Starter observe, Solo/Pro inject, Professional/Team enforce)
- Compliance frameworks plug into MOE PreToolUse compliance gate
v1.2.0
Shipped2026-05-19Adoption story: TIER 1 + TIER 3 commands · voice agents · 10 behavioral frameworks · full roster reorganization- TIER 1 verb-first commands (4 NEW): /mxm-build, /mxm-fix, /mxm-ship, /mxm-explain, plus /mxm-plan and /mxm-review upgraded to TIER 1 surface with Fogg B=MAP + conditional auto-loops
- TIER 3 persona dispatchers (5 NEW): /mxm-legal, /mxm-arch (native Wardley Mapping), /mxm-secure (triple-OWASP + AIBOM + NIST AI RMF + MITRE ATLAS), /mxm-founder (Prospect-Theory pricing), /mxm-pm (Jobs Atlas): 28 sub-commands handled inline
- Voice writing agents per ADR-016: nk-writer routes every writing task through myVoiceDNA/VOICE_SELECTION.md (22 content types, 15K token cap)
- 9-mode /mxm-help with persona auto-detect from project-manifest + cache at .mxm-skills/operator-persona.txt
- 10 NEW behavioral frameworks at ADR-007 7-section depth: TTM, SDT, Dual Process, Prospect Theory (HIGH); SCARF, Theory of Planned Behavior, Social Learning, Operant Conditioning, Diffusion of Innovations, Emotional Design (MED). 64 → 74 total.
- Full WS5 roster reorganization: 19 deprecations + 19 net-new = 91 agents (exact AGENT_ROSTER_v1.2_PROPOSAL.md target). CSO 9→19 (10 new specialists incl. gdpr-counsel, hipaa-counsel, llm-security-specialist, sbom-analyst, dpia-specialist). CINO 4→8. Orchestrators 5→10 (5 new structural-enforcement agents)
- Proactive Watch 11 → 13 drift classes: Class 12 behavioral-moat-drift ratified + Class 13 third-party-plugin-drift codified per ADR-012 MOE
- Capability deltas v1.1.1 → v1.2.0: agents 90→91 · skills 34→35 · commands 39→48 · frameworks 64→74 · drift classes 11→13
v1.2.1.0
Shipped2026-05-20NotebookLM research synthesis: the 9th MCP server- ADR-018 external-tool integration pattern: NotebookLM ships first-class via the 9th MCP server mxm-notebooklm (38 tools), with multi-office routing (CINO research · CMO content · CPO learning artifacts)
- CSO compliance-orchestrator auto-loop scans sources for PII / regulated data before Google ingestion
v1.3.0 → v1.3.1
Shipped2026-05-20Multi-tenant readiness: behavioral install wizard + per-operator voice lock- ADR-019: tier-aware install wizard with behavioral framings (loss aversion + default effect + endowment effect) reduces 14 install commands to 1 decision
- 90-day Trial of all 14 packs defaults at install: operators see the moat on real work before choosing a tier
- Per-operator voice-writer template instantiated via /mxm-brand-voice calibrate
v1.3.2 → v1.3.7
Shipped2026-06-20Default-on router · bounded loops · 78 frameworks · invisible updates- Maxim Default-On (ADR-021): a prompt-submit router routes every prompt to the right executive office automatically and shows the routing token cost up front (Claude Code CLI)
- loops skill: bounded agent-loop orchestration; added 4 loop-derived behavioral frameworks (74 → 78)
- awesome-agent-skills fallback registry: 1000+ external skills searched when no native skill matches (flagged Maxim-UNENHANCED per ADR-008)
- v1.3.7: invisible MCP dependencies on update: node_modules reused offline, so updates never cost a slow first restart
v1.3.8 → v1.3.8.2
Shipped2026-06-26Autonomous Workflow Standard + 14 everyday skills- Autonomous Workflow Standard (ADR-022) + the mxm-orchestrator engine: unattended workflows with a budget guard (hard-kill on token / tool / runtime / cost), a verification gate that runs separately from generation, idempotency, and dry-run ON by default. New /mxm-workflow command + orchestrator skill (29/29 acceptance)
- 14 everyday skills (benefit-first surfaces): quote-engine, inbox-triage, daily-digest, competitor-watch + 10 more; the connector-consuming ones govern existing Claude connectors (Gmail / news / web) rather than rebuild them
- v1.3.8.1: count/version tooling repair + fail-closed pre-commit drift gate · v1.3.8.2: commercial-pack loading fix + the 49th command (/mxm-workflow) on Desktop/Web
- Current scope: 91 agents · 52 skills · 49 commands · 9 MCPs (95 tools) · 78 frameworks · 14 compliance · 13 drift classes · 16 hooks · 21 ADRs
Planned: AI Governance + RevOps
PlannedTBDAI Governance + Revenue Operations- 5 AI governance frameworks shipped: NIST AI RMF, OWASP LLM Top 10, MITRE ATLAS, Constitutional AI, AI Bill of Materials (AIBOM / SPDX 3.0)
- New Revenue Operations domain under CMO: pipeline health, churn prevention, form CRO, sales engineering, customer success, SaaS metrics, A/B experimentation
- Behavioral framework pairing: Prospect Theory for churn-signal loss framing, Diffusion of Innovations for adoption-curve messaging, EAST for CRO nudges
v1.4
PlannedQ2 2027Vertical operator rosters (CRO + Fintech)- Regulated Industries operator roster: FDA submissions, MDR technical files, CAPA workflows, ISO 13485/14971 QMS, ISMS 27001 audits
- Fintech Specialist domain: payment integration, quant analysis, risk modeling, PCI-grade flows, investment & financial modeling
- Healthcare / Legal / Fintech / GovTech overlays gain operator depth: compliance-awareness evolves into compliance-authorship
v1.5
PlannedQ3 2027Enterprise + multi-seat + Cinematic / Video-AI- Agency tier: $599/mo, 20 seats, multi-tenant, priority support, overlay discounts
- Enterprise tier: custom pricing, SSO, SLA, on-prem option
- 16 IDE adapter integrations (VS Code, JetBrains, Cursor, Zed, and more)
- LemonSqueezy as secondary payment processor for EU VAT handling
- Team audit trail export (SOC 2 evidence pack)
- Cinematic / Video-AI Production domain: shot language, motion design, moodboard synthesis, recipe libraries (Higgsfield pack already bundled, domain wraps it)
- MOAT_TRACKER public dashboard: live moat-row count + pack activation metrics
v2.0+
FutureCross-provider + beyond-Claude- Cross-provider routing: OpenAI, Gemini, Mistral, and local models via LiteLLM-style proxy
- Provider-aware cost optimization: cheapest model that still clears the confidence rubric
- Self-hosted pack-engine option for air-gapped environments
- Maxim agent framework SDK: build third-party L4 packs with the same behavioral moat contract
- Multi-language brand-foundation overlays (voice profile per locale)
- Proprietary ML-driven framework scoring (replaces v1.0 regex heuristics in mxm-behavioral)
How to read this roadmap
Planned items are expected within two release cycles. Future items are directional and subject to customer feedback, framework research availability, and technical feasibility. Frameworks expansion is governed by ADR-007 Behavioral Moat Framing Doctrine: no framework ships without an anti-pattern registry and a pack integration.
The detailed framework-by-framework roadmap lives in documents/reference/FRAMEWORK_ROADMAP.md in the public repo. This page is the release-level summary.