Changelog
What's new with HUMMBL Base120
HUMMBL
Changelog
Apr 14
2026
Python SDK docs + Quickstart guide
Docs
- hummbl-governance SDK reference — full API docs for all six primitives: DelegationTokenManager, BusWriter, KillSwitch, CircuitBreaker, CapabilityFence, MCP Attestation. Constructor signatures, method tables, and copy-paste examples. docs/hummbl-governance →
- hummbl-base120 SDK reference — Engine.select/apply/recommend, VERUM ledger API, CLI commands, MCP server setup. docs/hummbl-base120 →
- Quickstart guide — zero to governed AI agent in 5 minutes. Full copy-paste working example wiring kill switch, delegation tokens, governance bus, and circuit breaker. docs/quickstart →
- Site-wide JSON-LD repaired across 44 pages; COOP/CORP security headers promoted to global scope.
Apr 14
2026
base120 v2 SDK + hummbl-governance v0.4.0
SDK
-
base120 v2 SDK shipped —
pip install hummbl-base120. Stdlib-only Python package with 148 tests, 100/100 arbiter score. Includes Engine.select(), VERUM audit ledger, MCP server, and CLI. - hummbl-governance v0.4.0 — unified failure mode taxonomy: 30 FailureMode codes (FM1–FM30), 14 HummblError codes, and backward-compatible IDP_E_* aliases. 546 tests, 99.5/100 score.
- Governance CI hardened — removed 6 v1 leftover workflows and 19 orphaned corpus snapshots across both repos.
Feb 24
2026
Website Expansion
Website
- Added Playground page with live API sandbox, curated demos, safety stack visualization, and 66-agent lineup
- Added Explorer page for browsing and searching all 120 models
- Added Docs, Pricing, Case Studies, and Changelog pages
- Favicon, Open Graph images, and social sharing meta tags
- Mobile-responsive hamburger navigation across all pages
- Cloudflare Pages CSP headers for all routes
Feb 20
2026
Multi-Agent Platform
Platform
- 6 vendor runtimes: Claude (Anthropic), Codex (OpenAI), Kimi (Moonshot), Gemini (Google), Ollama (local), plus vendor-agnostic orchestration
- 60+ coordinated agents across the platform
- TSV append-only coordination bus for agent communication
- Agent registry with real-time status tracking
Feb 15
2026
Security Stack
Security
- 5-layer input/output security pipeline
- Prompt injection detection (18 patterns)
- PII detection and redaction (8 categories: SSN, credit cards, email, phone, API keys, passwords, tokens, private keys)
- Input sanitization with comment stripping and template neutralization
- Output validation with exfiltration detection
- Security event logging and monitoring via /health endpoint
Feb 10
2026
MCP Server
Integration
- Published
@hummbl/mcp-serveron npm - Use Base120 models directly from Claude Desktop, Cursor, Windsurf, or any MCP-compatible client
- Tools: recommend_models, list_models, get_model, match_workflows
- Zero configuration — just add to your MCP config
Feb 5
2026
Workflows Engine
API
- Added
/v1/workflows/matchendpoint - Pre-built workflows: Team Performance, Root Cause Analysis, Crisis Response, Strategic Planning
- Workflows chain multiple models into step-by-step reasoning sequences
- Pattern matching connects problems to relevant workflows automatically
Jan 28
2026
API Launch
Launch
- HUMMBL Base120 API goes live on Cloudflare Workers
- 120 mental models across 6 transformations
- Core endpoints: /v1/models, /v1/models/:code, /v1/recommend
- Rate limiting: 100 requests/minute per IP
- Global edge deployment via Cloudflare
- Zero authentication required during beta
Jan 15
2026
The Beginning
Announcement
- HUMMBL project founded
- Base120 framework designed: 120 mental models organized into 6 transformations (Perspective, Inversion, Composition, Decomposition, Recursion, Synthesis)
- Mission: Give AI agents structured thinking tools, not just raw intelligence