The Primitives

Governance infrastructure,
as libraries you import.

Seven load-bearing primitives. Each one is a Python package. Each one is stdlib-only. Each one is open source. Import what you need, own the policy.

Doctrine terms are anchored in the HUMMBL Glossary: Base120, BaseN, HUAOMP, and MTSMU, with unified definitions reused across docs, Method, and the architecture map.

$pip install hummbl-governance
These are not dashboards. They are the machinery dashboards describe.
Audit

Governance Bus

Append-only JSONL audit log. Assessor-readable, grep-friendly, evidence-preservation compliant. 

bus.write(
  actor="agent-1",
  action="deploy",
  scope="prod"
)
Learn more → Safety

Kill Switch

Four-mode emergency halt. Runtime-enforced, file-system-persisted, survives process restart. 

ks.engage(
  mode=HALT_ALL,
  reason="anomaly"
)
Learn more → Resilience

Circuit Breaker

CLOSED/HALF_OPEN/OPEN state machine wrapping external adapters. Fails fast when upstream degrades. 

cb.call(
  adapter,
  timeout=30
)
Learn more → Identity

Delegation Tokens

HMAC-SHA256 signed capability tokens. Runtime attribution for every agent action. 

dct = DCT.issue(
  scope="read:docs",
  depth=2
)
Learn more → Scope

Delegation Context

Chain-depth enforcement. Prevents privilege escalation via chained delegations. 

dctx = DCTX(
  parent=tok,
  max_depth=3
)
Learn more → Reasoning

Base120 Models

120 canonical mental models. Structured reasoning for agents. API, MCP server, source SDK. 

model = b120.get("P1")
# First Principles
Learn more → MCP

MCP Attestation

Local-first governance for MCP-native agents. Capability tokens, audit receipts, server allowlists. 

attest.verify(
  server="postmark",
  policy=ALLOWLIST
)
Learn more →