Capability Based Security Python Packages

aindy-runtime

Self-hosted runtime for AI agents and workflows. Durable execution, capability-gated tools, human approval gates, and a memory system that accumulates. Runs survive restarts and resume where they stopped. Mount your domain logic as a plugin package or drive it over the HTTP SDK

attenu-guard

Permissions that narrow when one AI agent delegates: a sub-agent never holds more than its parent. In-process, with a hash-chained audit log you verify offline. Python and TypeScript, adapters for 19 frameworks and A2A.

ephemora-cell

Ephemeral, stateless, capability-bound execution for untrusted AI and agent-generated code.

taper-broker

A credential broker for AI agents. The agent names a typed operation instead of holding a credential, and tokens narrow but never widen. Ed25519 capability chains, Postgres and SSH adapters, 231 red-team scenarios. Unaudited.

capability-reasoning-kernel

Python reference for a CaMeL-like Reasoning Kernel. Control/data-flow separation reduces prompt-injection-to-effect risk when planner input is trusted, the capability policy is correct, and tools enforce it. Topology, not a proof that injection cannot cause an unauthorized effect.

attenu-derive

Attenu Derive: computes the permission set each AI agent task needs from the app itself — roster, tools, observed calls — for attenu-guard to enforce on every tool call and sub-agent handoff.

kinegrant-protocol

Reference implementation of KineGrant, an authorization and accountability layer for physical AI

cicash

A budget you lend to an AI agent, not money you give it: bounded, expiring, revocable, auditable spending authority. Wire spec + two independent implementations.

weaver-kernel

Capability-based authorization and policy enforcement for agents using large MCP and A2A tool ecosystems.

quantumos-host-tools

A quantum-aware microkernel OS — capability-based IPC, real QPU entropy at boot, holographic kernel memory. Boots in your browser.

kernos-sdk

An open engine for company-wide LLM agent automation. Agents work inside a signed remit, stop at the policy gates you set, and keep running unattended for months.

niyam-sdk

Self-hosted AI governance in one binary: gate every model call, action, and data path against central policy, with a post-quantum-signed audit trail.

vac-client

VAC Protocol - Capability-based security for AI agents. Task-scoped credentials, receipt-based state, instant revocation.