Authority
You define the boundary.
Govern what AI agents can do. Prove what they did.
Set a boundary. Watch an action respect it. Get evidence you can inspect.
Open source. The Python package runs in your environment; this page is a browser walkthrough with fictional data.
Browser walkthrough fictional sample data
Allow reading project notes
Controls whether the browser walkthrough may read the fictional project notes.
project-notes.txt
Read the project notes.
This result reflects the previous policy setting. Run the action again to apply the new boundary.
project-notes.txt
Run the action to create a browser-session receipt.
Browser-session receipt
The page creates a non-exportable, ephemeral HMAC key in this browser session. Verification detects changes made after signing; it does not establish an independent identity, execute a model, or provide an operating-system sandbox. Use the Python package for native enforcement in your own execution path.
You define the boundary.
The action respects it.
The result leaves a record.
Source-linked evidence
The walkthrough above explains the interaction. The figures below describe the cited public package and repository evidence snapshot.
Snapshot scope: the cited v1.4.2 package is classified Alpha. Public repository CI reported 2,463 passed, 3 skipped. These measurements describe the linked source and run, not a general performance or suitability claim.
Bind an action to declared operations, resources, and task scope.
Place the policy decision in the execution path it governs.
Record the decision and verify the record inside its trust domain.
This example runs against the package API and composes authorize, execute, record, bind, and verify explicitly.
import hashlib
import secrets
from pathlib import Path
from tempfile import TemporaryDirectory
from hummbl_governance import DelegationTokenManager, ReceiptEngine
secret = secrets.token_bytes(32) # Load a stable key in production.
contract = {"id": "summary-v1", "operations": ["summarize"]}
task_id = "task-42"
# AUTHORIZE — issue and validate scoped delegated authority.
tokens = DelegationTokenManager(secret)
dct = tokens.issue(
issuer="orchestrator",
subject="summarizer-agent",
operations=contract["operations"],
resources=["docs/report.md"],
task_id=task_id,
contract_id=contract["id"],
)
allowed, reason = tokens.validate_token(
dct,
expected_task_id=task_id,
expected_contract_id=contract["id"],
expected_subject="summarizer-agent",
)
if not allowed:
raise PermissionError(f"Delegated authority denied: {reason}")
# EXECUTE — your runtime performs the authorized action.
source = "Bounded authority makes agent actions reviewable."
result = " ".join(source.split()[:5])
# RECORD + BIND + VERIFY — authenticate the tuple and verify its chain.
with TemporaryDirectory() as state:
receipts = ReceiptEngine(Path(state), signing_secret=secret)
receipt = receipts.create_and_store(
agent_id=dct.subject,
action_type="summarize",
payload={
"contract": contract,
"dct": dct.to_dict() | {"signature": dct.signature},
"evidence": {
"outcome": "completed",
"output_sha256": hashlib.sha256(result.encode()).hexdigest(),
},
},
)
if not receipts.validate(receipt):
raise RuntimeError("Receipt authentication failed")
chain_valid, chain_reason = receipts.verify_chain(dct.subject)
if not chain_valid:
raise RuntimeError(f"Receipt chain failed: {chain_reason}")
if len(receipts.list_for_agent(dct.subject)) != 1:
raise RuntimeError("Receipt persistence check failed")
Version 1.4.2 does not expose a dedicated GovernanceTuple object or a one-call lifecycle. The application composes the tuple in the authenticated receipt payload.
HMAC-SHA256 delegation tokens bind issuer, subject, task, and authority fields inside a shared-secret trust domain. HMAC receipts make integrity and authenticity checkable by parties that hold the shared key. HMAC verifies integrity and authenticity inside that boundary; it is not public-key proof.
The published package is Alpha. A live public GET on https://hummbl-receipt-probe.hummbl.workers.dev/ produced 2026-08-31T21:56:55Z (receipt r-7e400da03299). That scoped event does not support a general “production-tested” claim.
A primitive only mediates actions routed through it. The implementer remains responsible for complete integration, key protection, identity, evidence capture, monitoring, and response.
HUMMBL supports technical evidence generation and engineering mappings. It does not determine legal applicability, confer compliance, certify an AI management system, or replace an auditor, standards body, counsel, or regulator.
EU AI Act transparency and GPAI enforcement now apply. Later high-risk milestones should be evaluated against authoritative guidance for the deploying organization.
Review every claim, source, scope, and limitation in the public claim ledger.
Authority, containment, and evidence. Open tools for agents that act on your terms.