@agent-lint/mcp

Read-only MCP server for keeping AGENTS.md, CLAUDE.md, rules, skills, workflows, and plans structured, current, and codebase-aware.

Start the Server

npx -y @agent-lint/mcp

For most users, the fastest setup path is still:

npx @agent-lint/cli init

Current Tool Surface

| Tool | Purpose | | --- | --- | | agentlint_get_guidelines | Return artifact guidance before creating or updating context files | | agentlint_plan_workspace_autofix | Scan a workspace and return a step-by-step fix plan | | agentlint_quick_check | Decide whether recent code changes require context updates | | agentlint_emit_maintenance_snippet | Return a reusable maintenance snippet for managed client files or AGENTS.md / CLAUDE.md fallbacks |

Current Resource Surface

| Resource | Purpose | | --- | --- | | agentlint://guidelines/{type} | Readable guidelines for one artifact type | | agentlint://template/{type} | Skeleton template for a new artifact | | agentlint://path-hints/{type} | IDE-specific file discovery hints |

Programmatic Usage

import { createAgentLintMcpServer } from "@agent-lint/mcp";

const server = createAgentLintMcpServer({
  transportMode: "stdio",
});

HTTP Transport

npx @agent-lint/mcp --http --port 3001

Notes:

• HTTP mode keeps the public surface read-only.
• Workspace scanning is disabled in HTTP mode unless you explicitly enable it.

Design Constraints

• No server-side LLM calls
• No database or cache
• No file writes from the MCP server
• Local stderr logging only

Related

• Root README
• CLI package
• Authoritative publish source

License

MIT