Intent
What action is being proposed, which objective it serves, where it originated, and whether the intent is explicit enough for the operation.
AGEC / Public Alpha
A governance gate between agent reasoning and execution.
AGEC evaluates whether a proposed AI agent action should proceed at the moment of execution, using its intent, runtime context, and planned execution path.
It returns a structured decision - proceed, review, suspend, halt, or reauthorize - and records the reason for audit.
Built for teams deploying agents that invoke tools, modify systems, send communications, or create other external side effects.
Run the CLI demo: agec-demo
v0.2.0 Public Alpha
Python 3.10+
Apache-2.0
Available on PyPI
Suitable for local evaluation, demos, and early pilot integrations.
Execution Gate
reauthorize
Intent
send_price_list
source: user_request / confidence 0.91
Context
price_list_status: current
campaign_status: active
Execution Path
price_campaign_v2
unregistered sequence
Reason
The proposed execution path is not registered for this action.
audit_id
audit_7e51c09a22f1
The Operating Concept
Legitimate Execution Governance
We use Legitimate Execution Governance to describe the runtime evaluation of whether a proposed agent action remains valid at the exact moment it is about to affect an external system.
AGEC expresses this evaluation as a structured pre-execution decision based on intent, runtime context, and execution path.
Context
Are the facts required for the action present, current, and consistent with the declared execution conditions?
Execution Path
Is the agent following the registered sequence of steps and required validations for this action?
How It Works
One decision before the tool runs.
Describe the proposed action
Provide the action intent, the runtime facts required for execution, and the planned sequence of tool calls.
Evaluate at the boundary
AGEC validates the inputs immediately before the tool, node, or callable is invoked.
Apply the decision
Execution continues only on proceed. Other decisions
can trigger review, suspension, termination, or reauthorization.
Record the result
Each validation produces a reason and audit identifier that can be persisted for later inspection.
Where AGEC Sits
AGEC complements the systems around your agent.
AGEC complements identity systems, access controls, policy engines, and agent frameworks by evaluating the specific proposed action at the final pre-execution boundary.
Identity & Permissions
Agent Reasoning & Planning
AGEC Validation
Tool Execution
Audit
AGEC SDK
A pre-execution governance SDK for developers.
AGEC - Authorized Governance Execution Context - runs immediately before a tool, node, or callable and returns a structured governance decision that determines whether execution can continue.
It can guard regular Python callables, OpenAI Agents tool functions, and LangGraph-style nodes.
pip install agecfrom agec import AGEC, Intent, Context, ExecutionPath
decision = AGEC().validate(
intent=Intent(
type="send_price_list",
source="user_request",
confidence=0.91,
),
context=Context(
facts={"price_list_status": "current"}
),
execution_path=ExecutionPath(
steps=[
"crm.read_customers",
"pricing.get_latest_list",
"email.send_campaign",
],
approved_path_id="price_campaign_v1",
),
)
print(decision.status)Run the deterministic demo with agec-demo.
SDK Capabilities
Small enough to evaluate, explicit enough to trust.
Public Python SDK
Install the current alpha from PyPI and place the gate where your tools execute.
Explicit Decisions
The current SDK produces five defined outcomes: proceed, review, suspend, halt, and reauthorize.
Auditable by Design
Every decision includes a reason and audit ID. Audit events can be persisted as JSONL.
Framework Adapters
Guard regular callables, OpenAI Agents tool functions, and LangGraph-style nodes without handing AGEC ownership of your framework clients or API keys.
Use Case
A price campaign before execution
A sales agent prepares a campaign to send an updated price list to a customer segment. Immediately before the email tool runs, AGEC evaluates the campaign intent, checks the current runtime facts, and compares the planned tool sequence with the registered execution path.
proceed
A current price list and registered path can return proceed.
review
Missing runtime facts can return review.
suspend
An invalid context can return suspend.
halt / reauthorize
A changed execution path can return halt or reauthorize.
The decision is made before any campaign email is sent.
Planoryx
AGEC is the first product developed by Planoryx.
Planoryx develops the SDK, validation services, review workflows, and observability required to operate governed AI agents across real business systems.
AGEC SDK - Available
The public Python SDK that evaluates proposed actions before tools execute.
Validation Service - Planned
Centralized validation, registered execution paths, and persistent decision records.
Review Operations - Planned
Queues and approval workflows for decisions that require human review or reauthorization.
Audit & Observability - Planned
Searchable execution history, governance metrics, and enterprise monitoring integrations.
Roadmap
From public alpha to operational governance.
Planoryx develops AGEC and the operational tooling around pre-execution governance. The roadmap is expressed as product maturity stages, not calendar promises.
-
Current
AGEC SDK 0.2 Public Alpha
Python package, five governance decisions, audit logs, CLI demos, callable wrappers, and OpenAI/LangGraph helpers.
-
Next
AGEC Validation Service
Centralized registered-path management, remote validation, persistent audit records, and MCP integration.
-
Later
Planoryx Operations Console
Review queues, approval workflows, governance dashboards, templates, and enterprise integrations.
Evaluate AGEC
Evaluate AGEC in your own workflow.
Install the public alpha, run the included demo, and place the governance gate before a tool or callable in your agent workflow.