agent-permissions

Three packages: @kernel.chat/agent-os (POSIX for AI agents — capabilities, namespaces, quotas, taint, audit, vault, outcomes), @kernel.chat/kbot (terminal AI agent, MCP-native, BYOK), @kernel.chat/kbot-finance (audit-grade AI for regulated industries). Provenance-engineering substrate.

On-chain spending policies for AI agent wallets. 14 testnets + 2 mainnets. 4 audits. ASP-1 spec.

Policy enforcement for AI agents, centralized and consistent. Define your rules once in umbod.toml, hook in your agents, and every one of them plays by the same rulebook.

Gate AI-agent actions with explicit grants, durable journals, MCP server support, and verification.

WRIT is a GenLayer project for agent permissions

MCP server for Agent Ready — scan any URL for AI agent readability against the Vercel Agent Readability Spec, llmstxt.org, and agent protocols (MCP, A2A, agents.json, agent-permissions, UCP, x402). 59 checks with per-check fix guidance.

Zero-knowledge authorization for autonomous AI agents. Cryptographically verify agent mandates without exposing policy details or sensitive authorization data.

A small offline reference agent with a narrow permission boundary and deterministic tests.

**The data behind agent permission rules.** Verified facts about what each CLI command does — destructive flags, network reach, BSD vs GNU differences — with a re-runnable check behind every row, so a policy is written from evidence instead of intuition.

Free AI agent authority, permissions and reachability resources for understanding what agents can access before they act.

POSIX for AI agents — ten OS-level primitives. Mirrored from isaacsight/kernel monorepo. Apache 2.0.

Permission manifest schema, validator, and evaluator for AI agents.

Public-safe security and boundary patterns for artifact publication, repo visibility, secrets handling, and data separation.