cedar-policy

Runtime enforcement boundary for AI agents: a local sidecar that gates every outbound call against Cedar policies you own. Deterministic, call-level, no model on the hot path

Open-source infrastructure for groups of AI agents — identity, capability, accountability, and norms. Framework-agnostic.

MCP server for Cedar policy language - validate, authorize, diff, and plan policy changes for Amazon Verified Permissions from your AI assistant.

Agent Run Config — an open specification for declaring, packaging, securing, and sharing portable, governed AI agents. Like a Dockerfile for agents: one reviewable Agentfile for identity, tools, boundaries, policy, and OCI packaging.

Apache-2.0 licensed lightweight agent sandbox: Cedar policy + Ed25519-signed receipts in one Rust binary. Design-partner preview.

Cedar for .NET — a C#/.NET port of the Cedar Java bindings, enabling .NET applications to parse, validate, format, convert, and evaluate Cedar policies using the native Cedar engine.

Deterministic, policy-as-code authorization for what an AI agent is allowed to do. Every tool call checked before it runs, logged after. A runnable reference build, not output guardrails. https://demo.sarthak-gupta.com

Open-source AI agent governance kernel — cryptographic audit trails, consent-checked data access, and verifiable decision records.

Policy platform for AI agents: Cedar policy over network, filesystem, and process access. Container sandboxing with observe-then-enforce workflow.

The Zero-Trust Action Hub is a standalone, Zero-Trust Policy Decision Point (PDP) designed for autonomous AI agent ecosystems. It enforces cryptographic governance over high-risk agent actions using AWS Cedar policies and Ed25519 digital signatures, requiring agents to collect and present cryptographic proofs from trusted external microservices.

Zero-trust API firewall and security integrity layer for autonomous AI agents & Model Context Protocol (MCP) tool execution. Secure, TOCTOU-proof, fail-closed.

PaediatricClinic Spring Boot application to demonstrate Cedar.

Experimental trust and governance ideas for autonomous AI agents — captcha, permit, mesh, eval, memory.

OpenAgentTrustStack (OATS) Specification

Open-source Identity Governance & Administration (IGA) platform. Graph-based (Neo4j), durable workflows (Temporal), Cedar policy-as-code. Humans + AI agents + IoT devices — one policy engine.

Reusable Cedar policy templates for agent action receipt verification. Interoperable with the Cedar policy engine and cedar for agents.

Multiplayer agentic coding: a Discord-like shared workspace where people and their AI agents (Claude Code, Codex) build software together — live shared timeline, steerable agents, E2E-encrypted channels, approval-gated cross-machine dispatch, git-worktree flow. Local-first: every agent runs on its owner's machine, on its owner's subscription.

Vendor-neutral authorization for AI agents. Run the check inside the agent controls you already use (LlamaFirewall, NeMo Guardrails, FastMCP, A2A) and decide with the Policy-as-Code engine you already trust (OPA, Cedar, OpenFGA, or any AuthZEN PDP). One fail-closed allow/block/human-review verdict on every agent action.

Reference architecture for governed agentic AI on AWS. Purpose-bound Cedar authorization at column level, Iceberg snapshot-pinned queries for replayable audit trails, and confidence-gated document extraction with Bedrock Data Automation and AgentCore, demonstrated on insurance claims processing with an EU AI Act mapping.

Kernel level behavioral security and adaptive memory layer for autonomous coding agents.