agent-governance (Python client)
Governance for AI agent tools — deny by default, L0–L4 guardrails, human approvals, full audit chain — via the agent-governance sidecar. Zero dependencies (stdlib only).
Install & adopt in three lines
# from this repository (works today, no PyPI needed):
pip install "git+https://github.com/NexusClawHQ/nexusclaw.git#subdirectory=governance/adapters/python"
# or from PyPI once published: pip install agentgovernance
from agent_governance import GovernanceClient
gov = GovernanceClient("http://127.0.0.1:7899")
update_customer = gov.wrap_tool(update_customer) # gated + audited
Every wrapped call now:
- asks the sidecar
POST /gatewhether the tool may run — grants and risk rules live server-side (deny by default: a tool nobody granted never runs, and the denial itself lands on the audit chain); - on
allow, executes locally and reports the outcome (complete), so the audit chain records the result; - on
blocked, raisesGovernanceDenied(reason); - on L2/L3 risk, raises
GovernancePendingApproval— wire it to your framework's human-in-the-loop.
LangGraph / CrewAI: the interrupt pattern
@gov.wrap_tool(name="demo.send_followup_email")
def send_email(customer_id: str, subject: str): ...
try:
send_email("C-1001", "quarterly check-in")
except GovernancePendingApproval as pending:
interrupt({"approval_id": pending.approval_id, "risk": pending.risk_level})
# ... after the human decides (console, API, anywhere):
gov.decide(pending.approval_id, "APPROVED")
result = gov.run_approved(send_email.__wrapped__, pending, "C-1001", "quarterly check-in")
Prefer to just block? wrap_tool(fn, wait=True) polls until a human decides
(approved executes + completes; rejected raises GovernanceDenied).
Running the sidecar
cd governance/packages/sidecar
SIDECAR_GATE_ALLOWED_TOOLS="crm.update_customer,demo.send_followup_email" \
SIDECAR_PGDATABASE=nexusclaw_sidecar pnpm exec tsx scripts/dev-server.ts
The console at http://127.0.0.1:7899/console shows pending approvals and
the audit chain (execution → tool calls → outbox events).
Tests
python3 -m unittest discover -s test # offline (stubbed transport)
python3 scripts/integration_test.py http://127.0.0.1:7899 # live sidecar
Metadata
Release files for nexusclaw-agent-governance 0.1.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| nexusclaw_agent_governance-0.1.0.tar.gz | 7.0 kB | Without provenance | Details for nexusclaw_agent_governance-0.1.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| nexusclaw_agent_governance-0.1.0-py3-none-any.whl | Python 3 | none | any | Details for nexusclaw_agent_governance-0.1.0-py3-none-any.whl |
Total release size: 13.5 kB
Release files / nexusclaw_agent_governance-0.1.0.tar.gz
| Download URL | nexusclaw_agent_governance-0.1.0.tar.gz |
|---|---|
| Size | 7.0 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
450b154fe459d5ebd45de437126e444eb46fd7f60da48668239f94d8f26dd792
|
|
BLAKE2b-256 checksum How to use checksums |
d727185de1cea9f6356a2b769e3e917cec7b68b262f1013bb0faed0cbc040111
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.14.5
|
Release files / nexusclaw_agent_governance-0.1.0-py3-none-any.whl
| Download URL | nexusclaw_agent_governance-0.1.0-py3-none-any.whl |
|---|---|
| Size | 6.5 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
3fbce8e1aaeb049b1b4dd75ba3bd7299361c8d7aa4c875d07061043ea7bbc21a
|
|
BLAKE2b-256 checksum How to use checksums |
550828af383a1fbc2d65b989c70afb14a516779498b08e81b065c0af838519ef
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.14.5
|
Security for nexusclaw-agent-governance 0.1.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.