@plasius/ai-mcp

MCP tool registry, per-call allowlist, and tool audit contracts for Plasius AI orchestration.

Scope

This package is part of the layered @plasius/ai-* package family. It defines the external contracts for MCP tool registration, per-call allowlisting, role-gated risk decisions, and audit metadata.

Install

npm install @plasius/ai-mcp

Contracts

• AI_MCP_FEATURE_FLAGS declares the feature flags that gate MCP behavior.
• resolveAiMcpToolAllowlist evaluates requested tool IDs against registry descriptors, feature flags, and actor roles.
• isAiMcpToolAllowed and isAiMcpToolRiskAllowed provide lightweight policy predicates for callers.
• packageDescriptor exposes package name, primary flag, env prefix, and summary.

Usage

import {
  AI_MCP_FEATURE_FLAGS,
  resolveAiMcpToolAllowlist,
} from "@plasius/ai-mcp";

const result = resolveAiMcpToolAllowlist({
  requestedTools: ["rag.search", "admin.kill-switch"],
  actorRole: "operator",
  featureFlags: {
    [AI_MCP_FEATURE_FLAGS.mcp]: true,
  },
  toolRegistry: [
    {
      toolId: "rag.search",
      toolName: "RAG Search",
      riskClass: "safe",
    },
    {
      toolId: "admin.kill-switch",
      toolName: "Admin Kill Switch",
      riskClass: "restricted",
    },
  ],
});

console.log(result.allowedTools);
console.log(result.blockedTools);

Development

npm install
npm run build
npm test
npm run test:coverage
npm run pack:check

Release Workflow

Protected main releases use a two-step flow:

1. Run .github/workflows/cd.yml with bump=patch|minor|major to open or refresh a release/vX.Y.Z prep PR.
2. Merge that PR to main.
3. Rerun .github/workflows/cd.yml on main with bump=none to tag, draft the GitHub release, and publish to npm.

Governance

• Security policy: SECURITY.md
• Code of conduct: CODE_OF_CONDUCT.md
• ADRs: docs/adrs
• CLA and legal docs: legal

License

Apache-2.0