Governance infrastructure for AI agents.
HUMMBL builds small, inspectable control-plane primitives for agentic systems: delegation tokens, append-only receipts, kill switches, circuit breakers, capability fences, source-verification gates, and review workflows that create evidence while the system runs.
The bias is explicit:
- libraries over platforms — controls should live in the execution path, not only in a dashboard after the fact
- receipts over vibes — every governance claim should leave an audit trail
- stdlib-first primitives — core safety controls should survive constrained, regulated, and air-gapped environments
- human override preserved — automation assists governance; it does not become the sovereign
pip install hummbl-governance
| Need | Repository |
|---|---|
| Runtime governance primitives | hummbl-governance |
| Mental-model substrate | base120 |
| Source-verification gate | evidence-gate |
| Code quality and governance scoring | arbiter |
| MCP access to HUMMBL models and skills | mcp-server |
| Agent orchestration patterns | hummbl-agent |
| Research and citation corpus | hummbl-bibliography |
| Public profile, tools, and repo inventory | hummbl-dev |
| Area | What is shipping |
|---|---|
| Agent governance | hummbl-governance, agent-governance-demo, hummbl-agent, mcp-server |
| Evidence and review | evidence-gate, arbiter, governed PR review protocols, source-verification rules |
| Cognitive substrate | base120, hummbl-theory, hummbl-bibliography, arcana |
| Operator infrastructure | founder-mode, claude-config, hummbl-iac |
| Product research | fractional-bench, coaching, hummbl-production |
| Experimental public repos | autoresearch, governed-compression, sint-protocol, bif |
Some repositories are private while they contain operator workflows, client-sensitive research, or active product work. Public repositories are the durable reference surface.
Governance is not a policy PDF. It is a runtime property:
- Who is acting? Agent identity, trust tier, and delegation scope.
- What authority do they have? Signed capability tokens and bounded tools.
- What happened? Append-only receipts, bus messages, and evidence logs.
- What should stop the system? Kill switches, circuit breakers, cost governors, and review gates.
- What can be audited later? Source manifests, PR reviews, test evidence, and structured findings.
That is the spine across the org: small primitives that make governance observable, testable, and reviewable.
Inventory captured on 2026-05-08:
| Metric | Count |
|---|---|
| Total repositories | 54 |
| Active public repositories | 35 |
| Active private repositories | 17 |
| Archived repositories | 2 |
| Primary language family | Python-heavy, with TypeScript, HTML, Shell, TeX, Go templates, Java |
Latest repo inventory: docs/GITHUB_REPO_INVENTORY_2026-05-08.md
Latest branch cleanup audit: docs/BRANCH_CLEANUP_AUDIT_2026-05-08.md
HUMMBL's public thesis is that AI governance will move from abstract policy to runtime evidence. The research corpus tracks:
- AI-generated code risk and liability
- agentic control-plane design
- source verification before publication
- governance receipts and audit trails
- mental-model infrastructure for agent reasoning
- practical compliance mapping across NIST AI RMF, ISO 42001, EU AI Act, OWASP, and related frameworks
Useful entry points:
- AI slop crisis research corpus
- Why libraries, not platforms
- Reasonable care in the age of AI agents
- The observability argument
Free self-assessments and references:
| Tool | Purpose |
|---|---|
| Tool index | Landing page for the public tools in this repo |
| EU AI Act readiness | Governance posture check |
| NIST AI RMF readiness | Framework-aligned assessment |
| ISO 42001 readiness | AI management system checklist |
| Singapore agentic AI readiness | Agent governance architecture assessment |
| Colorado AI Act readiness | State-level algorithmic decision compliance |
| Compliance calendar | AI governance timeline |
| Governance crosswalk | Framework comparison |
HUMMBL, LLC is founded by Reuben Bowlby in Atlanta, GA.
- Site: hummbl.io
- Email: reuben@hummbl.io
- Package:
hummbl-governanceon PyPI
Governed agents need receipts, not slogans.