@xemahq/tool-policy

Composable policy chain for an agent's effective tool set

Overview

A composable policy chain that computes the effective tool set of an agent session from run context. It replaces ad-hoc, scattered enable/disable flags with a single chokepoint: each input is one contributor, and the chain is a reduce that is easy to read and test. New inputs (feature flags, plan, RBAC, permissions) are added as new contributors. Pure types plus a small composer, with zero runtime-adapter knowledge.

When to use it

• Use it to derive an agent session's effective tools from layered policy inputs.

Installation

pnpm add @xemahq/tool-policy

Usage

import { composeToolPolicies, EMPTY_TOOL_CONFIG } from '@xemahq/tool-policy';

const resolve = composeToolPolicies(contributors);
const effective = resolve(context, EMPTY_TOOL_CONFIG);

License

Apache-2.0 © Xema — xema.dev