Cage-lite: Prebind assurance for consequential AI agent actions

Notifications You must be signed in to change notification settings

Fork 0

Star 0

BranchesTags

Open more actions menu

Folders and files

NameName

Last commit message

Last commit date

Latest commit

History

71 Commits

71 Commits

.github/workflows

.github/workflows

cage_lite

cage_lite

docs

docs

examples

examples

playground

playground

tests

tests

.gitignore

.gitignore

CHANGELOG.md

CHANGELOG.md

CODE_OF_CONDUCT.md

CODE_OF_CONDUCT.md

CONTRIBUTING.md

CONTRIBUTING.md

LICENSE

LICENSE

NOTICE

NOTICE

README.md

README.md

SECURITY.md

SECURITY.md

pyproject.toml

pyproject.toml

requirements.txt

requirements.txt

Repository files navigation

Prebind assurance for AI-agent actions at the business consequence boundary.

Try the live CAGE-lite demo →

Install from PyPI · View release v0.1.2

Agent platforms govern how agents run. CAGE governs whether agent actions are allowed to become business consequences.

CAGE-lite is my open-source reference implementation of CAGE: Control, Assurance, and Governance Evaluation.

The broader framework is introduced in CAGE-1. CAGE-lite focuses on making its Prebind Assurance model visible, testable, and easier to evaluate in practice.

The project started with a simple question:

An AI agent can propose an action, but what should happen before that action becomes real?

Before an agent releases a payment, grants access, approves a transaction, updates a system of record, or discloses protected information, an organization should be able to verify that the action is authorized and supported by the required evidence.

CAGE adds that final assurance step before the action becomes a binding business consequence.

CAGE-lite is currently a v1 product preview. The Python package version is 0.1.2, and the current CAGE Warrant schema is version 0.4.

CAGE-1 defines a broader evaluation framework across twelve areas. This release focuses on the Prebind Assurance path: evaluating an action before it crosses the business consequence boundary, recording the decision, and proving whether the resulting effect was blocked or allowed to bind.

Product preview

The CAGE-lite dashboard shows the latest boundary decision, the original held action, recent boundary runs, and the result of replaying the action after the missing evidence is supplied.

The hosted demo automatically loads the standard USD 75,000 held-to-admitted replay, allowing visitors to inspect the boundary decision, CAGE Warrants, replay linkage, and effect proof without installing anything locally.

Quick start

CAGE-lite requires Python 3.10 or later.

Install from PyPI

In an active Python 3.10 or later environment:

python -m pip install cage-lite python -m cage_lite.demo.payment_replay

This runs the packaged payment replay and writes the generated artifacts under playground/v04-replay-demo/ in the current working directory.

To run the Streamlit dashboard, inspect the examples, or contribute changes, install CAGE-lite from source.

Windows PowerShell

git clone https://github.com/roopamwsure/cage-lite.git cd .\cage-lite

python -m venv .venv .\.venv\Scripts\Activate.ps1

python -m pip install --upgrade pip python -m pip install -e .

python -m cage_lite.demo.payment_replay python -m streamlit run cage_lite/ui/app.py

macOS and Linux

git clone https://github.com/roopamwsure/cage-lite.git cd cage-lite

python3 -m venv .venv source .venv/bin/activate

python -m pip install --upgrade pip python -m pip install -e .

python -m cage_lite.demo.payment_replay python -m streamlit run cage_lite/ui/app.py

The replay demo creates one original HELD Warrant and one linked ADMITTED replay Warrant under:

playground/v04-replay-demo/

The Streamlit application loads those artifacts by default. Developer controls remain hidden unless they are explicitly enabled.

Examples

The examples/ directory contains smaller demonstrations of individual CAGE behaviors:

payment_policy_demo.py evaluates the payment policy and produces a held boundary decision without attempting an effect.

payment_no_bind_demo.py shows that a held action does not execute and records durable NO_BIND effect proof.

payment_approval_demo.py adds the required approval, admits the action, executes the protected effect, and records BOUND proof.

payment_narrowed_demo.py narrows the requested payment to the agent's permitted scope and records the scoped effect result.

Run an example from the repository root:

python examples/payment_no_bind_demo.py python examples/payment_approval_demo.py python examples/payment_narrowed_demo.py

The examples write local CAGE Warrants, evidence records, and effect records under playground/. The generated output is excluded from Git.

Where CAGE fits

CAGE does not replace agent runtimes, IAM, policy engines, guardrails, gateways, approval systems, observability platforms, or agent evaluation frameworks.

Those systems produce important signals. CAGE consumes those signals and evaluates whether a proposed action should be allowed to cross the business consequence boundary.

The diagram above shows the broader CAGE framework. CAGE-lite is the open-source implementation used to make this assurance model visible, testable, and easier to evaluate.

The basic idea

A simplified CAGE flow looks like this:

Agent proposes an action | v Identity, standing, policy, and approval signals | v CAGE prebind boundary | +---- HELD ----> NO_BIND ----> Business effect blocked | +---- ADMITTED -> BOUND ------> Business effect executed | v CAGE Warrant and effect proof

CAGE Warrant

Each evaluated action produces a CAGE Warrant containing decision proof, effect proof, evidence references, replay linkage, and integrity information.

The Warrant distinguishes between deciding that an action may proceed and proving what happened after that decision.

Held-to-admitted replay

The included demo begins with a USD 75,000 vendor payment that exceeds the agent's USD 50,000 direct standing limit.

Without the required human approval, CAGE holds the action before effect execution:

boundary: HELD;

effect: NO_BIND;

system of record: NOT_WRITTEN.

The action is then replayed after approval evidence is added. The action, amount, standing limit, and policy remain unchanged. Only the approval state changes.

The replay is admitted, the effect is allowed to bind, and the original held Warrant remains preserved and linked to the replay Warrant.

Project links

Live CAGE-lite demo

CAGE-lite on PyPI

CAGE-lite v0.1.2 release

CAGE-1 on arXiv

CAGE-1 author-hosted PDF

Changelog

Contributing guide

Security policy

Report an issue or request a feature

License

CAGE-lite is licensed under the Apache License 2.0.

Apache-2.0 license

Code of conduct

Code of conduct

Contributing

Contributing

Security policy

Security policy

Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks

Report repository