agent-runtime-governance 0.5.1

Framework-agnostic runtime governance for AI agents.

Agent Runtime Governance

A lightweight, framework-agnostic runtime governance framework for AI agents. It governs an immutable ExecutionContext through a deterministic middleware pipeline, then produces an explicit decision, an auditable execution, and a replayable trace.

Quick start

pip install agent-runtime-governance

from pathlib import Path

from agent_runtime_governance import ExecutionMode, Runtime

runtime = Runtime()
WORKSPACE_ROOT = Path.cwd().resolve()

@runtime.tool(execution_mode=ExecutionMode.READ_ONLY)
def read_file(path: str) -> str:
    candidate = (WORKSPACE_ROOT / path).resolve()
    try:
        candidate.relative_to(WORKSPACE_ROOT)
    except ValueError as exc:
        raise ValueError("path must remain inside the workspace") from exc
    return candidate.read_text(encoding="utf-8")

print(read_file("README.md"))

Add deterministic rules, semantic review, human decisions, and signed audit:

import os

from agent_runtime_governance import (
    ApprovalMiddleware, AuditMiddleware, HumanDecisionProvider,
    InvocationOptions, JSONLAuditSink, LLMMiddleware, RiskTier,
    Rule, RuleMiddleware, Runtime,
)

runtime = Runtime([
    RuleMiddleware([Rule("explicit-wipe", r"\bwipe\s+all\b", "bulk wipe is forbidden")]),
    LLMMiddleware(lambda ctx: True),
    ApprovalMiddleware(HumanDecisionProvider(lambda ctx, request: True)),
    AuditMiddleware(
        JSONLAuditSink("audit.log", sign_key=os.environ["ARG_AUDIT_HMAC_KEY"])
    ),
])

@runtime.tool(risk=RiskTier.HIGH, requires_approval=True)
def delete_file(path: str) -> bool:
    return True

delete_file(
    "old.log",
    _governance=InvocationOptions(input_text="remove the old application log"),
)

Why it exists

Approach	Deterministic boundary	Human decision	Audit and replay
Prompt-only guardrails	No	Ad hoc	No
Hand-written permission checks	Partial	Application-specific	Ad hoc
Agent Runtime Governance	Yes	Provider interface	Built in

Prompt instructions are useful guidance, but they are not an authorization boundary. This project places deterministic policy, explicit decisions, and audit outside the model while remaining independent from agent planning and model providers.

Architecture

Tool Registry
     |
     v
Immutable ExecutionContext
     |
     v
Runtime Pipeline
  Rule -> LLM -> Human Decision -> Audit
     |
     v
Explicit Decision -> Tool Executor -> Final Audit Snapshot

Gating middleware may stop execution and fails closed. Observing middleware cannot grant permission and normally cannot interrupt an allowed tool. An audit sink explicitly configured as critical is the exception: failed delivery stops the call because an unaudited privileged action is not allowed. Earlier denials cannot be overridden by later middleware.

Design principles

1. Deterministic first - code and policy establish the execution boundary.
2. Policy over prompt - prompts guide behavior; policies authorize actions.
3. Defense in depth - independent controls can only tighten a decision.
4. Framework agnostic - the runtime does not own planning or model calls.
5. Human in the loop - applications supply the decision channel.
6. Observability by default - each transition produces traceable state.
7. Immutable context - middleware returns a new context instead of mutating shared state.

Examples

• examples/standalone_demo.py: framework-free pipeline
• examples/cli_approval_demo.py: interactive human decision provider
• examples/langgraph_integration.py: LangGraph tool-node integration
• examples/openai_agents_integration.py: OpenAI Agents SDK tool integration
• examples/crewai_integration.py: CrewAI decorated tool
• examples/agno_integration.py: Agno function tool
• examples/llamaindex_integration.py: LlamaIndex FunctionTool
• examples/autogen_integration.py: Microsoft AutoGen FunctionTool
• scripts/replay.py: print the snapshots for a trace from JSONL audit data

Each framework adapter is intentionally an example rather than a runtime dependency. The governed function remains ordinary Python and can be wrapped by the framework's native tool interface.

Engineering controls

v0.2 adds immutable pipeline composition, lifecycle hooks, Python-native policy, metrics, retries, timeouts, and optional OpenTelemetry export:

from agent_runtime_governance import (
    MetricsMiddleware, Pipeline, PolicyMiddleware, RetryMiddleware,
    SimplePolicy, TimeoutMiddleware,
)

pipeline = Pipeline([
    PolicyMiddleware(SimplePolicy(admin_only={"restart_service"})),
    RetryMiddleware(max_attempts=2),
    TimeoutMiddleware(5.0),
    MetricsMiddleware(),
])
runtime = Runtime(pipeline)

@runtime.before_tool
def attach_region(ctx):
    return ctx.evolve(metadata={**ctx.metadata, "region": "cn-beijing"})

Pipeline edits return a new Pipeline; a live runtime is never mutated behind concurrent calls. Hooks may enrich context but cannot change status or decisions.

Policies, snapshots, and regression

Load a strict versioned policy with a deterministic digest:

from agent_runtime_governance import Runtime, YAMLPolicyLoader

document = YAMLPolicyLoader.load("examples/policy.yaml")
runtime = Runtime([document.middleware()])
print(document.version, document.digest)

Duplicate tool entries, unknown fields, invalid risks, and unsafe YAML tags are rejected. YAML remains a configuration format over the deliberately small SimplePolicy model; it is not a general policy language.

SnapshotMiddleware records immutable lifecycle snapshots. ReplayDebugger prints timelines and field-level diffs, while EvaluationSuite runs governance without executing tools. PolicyDriftDetector reapplies deterministic policy to the same recorded request identity and reports decision or risk changes.

python scripts/trace_debug.py snapshots.jsonl TRACE_ID
python scripts/trace_debug.py snapshots.jsonl TRACE_ID --diff 0 1
python scripts/trace_debug.py snapshots.jsonl TRACE_ID --mermaid

Plugins and integrations

Plugins register components during construction; the built runtime remains immutable:

from agent_runtime_governance import OPAClient, OPAPlugin, PluginManager

manager = PluginManager()
manager.load(OPAPlugin(OPAClient("http://localhost:8181", "agents/tools/allow")))
runtime = manager.build()

Third-party plugins can publish the Python entry-point group agent_runtime_governance.plugins. Entry points execute Python code and must be treated as trusted dependencies. The project does not download plugins or provide a marketplace.

Integration	Extra	Behavior
Prometheus	prometheus	Terminal status and duration; no trace/user labels
Slack	none	Denial/failure notifications; official HTTPS webhooks only
OPA	none	Minimal decision input; fail closed by default
LangGraph	langgraph	Governed function in a graph node
OpenAI Agents SDK	openai-agents	Governed async function tool
CrewAI	crewai	Decorated governed async tool
Agno	agno	Typed governed Python function
LlamaIndex	llamaindex	Governed FunctionTool async function
Microsoft AutoGen	autogen	Governed FunctionTool async function

Install only the integrations an application uses:

pip install "agent-runtime-governance[yaml,prometheus,crewai]"

Production reliability

v0.5 focuses on production behavior instead of adding a new agent framework:

• idempotent tool execution with mandatory keys and in-memory or SQLite stores;
• durable approval recovery with leased reserve/commit/release transitions;
• absolute deadline propagation through admission, middleware, identity, tools, and idempotency waits;
• cancellation handling that records mutating in-flight work as UNKNOWN;
• trusted HMAC identity envelopes with replay protection;
• contract validation for parameters, results, and payload size limits;
• reliable JSONL and SQLite audit sinks with hash-chain verification;
• bounded concurrency, external integration circuit breakers, and fault tests.

The production smoke suite starts real Docker services for OPA and the OpenTelemetry Collector, exports through OTLP HTTP, scrapes a real HTTP /metrics exposition endpoint, and can run a local Kind smoke with a pinned node image:

python integration/production_smoke.py --skip-kind
python integration/production_smoke.py

Production deployments must configure a trusted identity provider with require_verified_identity=True; caller-supplied user, tenant, and permissions fields are compatibility inputs, not a trust boundary. SQLite stores coordinate processes on one host and require a distributed adapter for multi-host deployments.

v0.5 release verification record

The release was validated on 2026-07-26 (UTC) on Windows 11 with Python 3.12, Docker Engine 29.4.2, and Kind 0.31.0, and on GitHub-hosted Linux runners:

• 359 tests passed locally with 88.52% branch coverage on Python 3.12;
• the same 359-test suite passed on Python 3.10, 3.11, 3.12, and 3.13 in GitHub Actions, with 88.63%-88.66% branch coverage; the enforced floor is 80%;
• 15 repository-policy tests passed;
• real OPA HTTP allow/deny, OTLP HTTP export to an OpenTelemetry Collector, Prometheus scraping, and a Kind 1.34.3 control-plane readiness check passed;
• the wheel and source distribution installed and imported from separate clean virtual environments; and
• an isolated environment with the OTel, YAML, and Prometheus extras had no known dependency vulnerabilities reported by pip-audit 2.10.1 at the time of the run; before publication, the root distribution was reported as unavailable on PyPI and was not treated as an audited third-party dependency.

These are point-in-time verification results, not a latency SLA or a guarantee against future advisories. CI repeats the test matrix, policy checks, dependency audit, and Docker integration smoke from clean runners.

Benchmarks measure incremental runtime overhead for baseline, Rule, OPA, Audit, OpenTelemetry, and 10-middleware pipelines:

python benchmarks/benchmark_runtime.py --requests 100,500,1000 --concurrency 100

The committed Windows/Python 3.12 measurement is available in benchmarks/results/v0.5.0-windows-python312.json. Its 100-waiter, zero-hold FIFO admission test measured 2.93 ms p99 wait time. This point-in-time result is regression evidence for that machine, not a cross-platform latency SLA.

Releases

Version	Scope
v0.1.0	Immutable context, registry, rule/LLM/approval/audit middleware, basic replay
v0.2.0	Hooks, Python policy, metrics, retry, timeout, and OpenTelemetry bridge
v0.3.0	Strict YAML policy, snapshots, replay diff, evaluation, and policy drift
v0.4.0	Trusted plugins, Prometheus, Slack, OPA, and six framework integrations
v0.4.1	Mandatory linked issues and merge-policy verification
v0.4.2	Fail-closed CodeRabbit review verification for the current commit
v0.5.0	Production reliability: idempotency, identity, durable approvals, audit, deadlines, cancellation, contracts, real integration smoke
v0.5.1	Security hardening: caller metadata isolation and exact approval binding

Released versions are preserved as immutable Git tags. See CHANGELOG.md for the detailed compatibility and security notes.

Non-goals

• Agent planning, prompting, model routing, or memory
• Multi-agent communication or distributed execution
• Approval UI or a hosted control plane
• A general policy language, plugin marketplace, or mutable live pipeline
• Production-grade time-travel debugging
• Arbitrary condition evaluation, policy inheritance, or conflict resolution

See ARCHITECTURE.md for invariants and ROADMAP.md for the deliberately staged scope. Security reports and integration boundaries are documented in SECURITY.md. Contributor workflow and plugin contribution rules are documented in CONTRIBUTING.md. Production configuration and recovery are covered in docs/production.md, and the release procedure is defined in RELEASING.md.

Development

python -m pip install -e ".[dev]"
pytest
python integration/production_smoke.py --skip-kind
python -m build

Python 3.10+ is supported. The core package depends on filelock and jsonschema; optional integrations are installed via extras.

Pull requests must link an existing issue in this repository with a closing keyword such as Fixes #123. A repository workflow closes unlinked pull requests automatically. Pull requests to main must also pass every CI job and the CodeRabbit status, receive a verified CodeRabbit approval for the current head commit, and resolve blocking reviews. A skipped, rate-limited, stale, or missing CodeRabbit review fails closed. GitHub does not allow a pull request author to approve their own change, so the single-maintainer phase has no human approval count; one code-owner approval and last-push approval become mandatory when a second maintainer is added. Maintainers still use the same issue and pull request flow as external contributors, and administrators cannot push directly to main or bypass protection.

License

MIT