feat(cli): configure MCP natively across coding agents

README.md

@@ -75,12 +75,36 @@ firecrawl setup skills
 firecrawl setup workflows
 ```
 
-To install the Firecrawl MCP server into your editors (Cursor, Claude Code, VS Code, etc.):
+To install the Firecrawl MCP server into your coding agents:
 
 ```bash
 firecrawl setup mcp
 ```
 
+This detects which agents you have installed, lists those in a picker
+(already selected), and asks whether to add rules telling those agents to
+prefer Firecrawl for web search and scraping. Supported agents are Claude Code,
+Cursor, VS Code, Codex, OpenCode, Hermes Agent, and OpenClaw.
+
+Setup writes to your global agent settings, so one command puts Firecrawl on
+every agent you already use. Pass agent flags to skip the picker, or `-y` to
+configure every detected agent (MCP only):
+
+```bash
+firecrawl setup mcp --claude --cursor    # skip the picker
+firecrawl setup mcp -y                   # every detected agent, MCP only
+firecrawl setup mcp -y --rules           # ...and install the rules too
+```
+
+Rerun the command any time to update an existing setup or add another agent; it
+edits only the Firecrawl entry and leaves the rest of each config alone.
+
+Your API key is never written into an agent config. When `FIRECRAWL_API_KEY` is
+exported in the environment your agents run under, each agent gets a reference
+to that variable in the syntax it understands. Otherwise setup stays keyless,
+which still serves search, scrape, and parse under an anonymous rate limit. Use
+`--keyless` to force the anonymous path even when a key is available.
+
 To make Firecrawl the default web provider for supported AI agents:
 
 ```bash

package.json

@@ -83,6 +83,7 @@
     "@inquirer/prompts": "^8.2.1",
     "commander": "^14.0.2",
     "firecrawl": "4.24.0",
+    "jsonc-parser": "3.3.1",
     "yaml": "^2.9.0",
     "zod-to-json-schema": "3.24.6"
   }