p202 mcp

Serve Prosper202 to AI assistants over MCP (stdio), for Claude Desktop, Claude Code, Cursor and others

New in 1.9.78
All commands

p202 CLI reference

Back to the docs

01Run it

Replace the <placeholders> with your own values, or use the builder below.

p202 mcp

02What you get

Pick the shape you need. People get a table. Add --json, --csv or --ndjson for scripts, or -q for ids only. An AI agent gets compact JSON without asking. All output formats

03Build your command

Pick values and the command line writes itself, quoted and ready to paste.

p202 mcp

Set flags below; the command updates as you type.

04Flags

1 flag, plus the global flags every command takes.

FlagWhat it does
--read-onlyboolServe only the tools that read; writes are refused

05How it works

Runs an MCP (Model Context Protocol) server on stdin/stdout. An AI app starts it and calls its tools, each of which runs one p202 command against your install with the profile that is active when the server starts (or --profile).

claude mcp add prosper202 -- p202 mcp --read-only (recommended to start) claude mcp add prosper202 -- p202 --staged mcp (writes become proposals) claude mcp add prosper202 -- p202 mcp

Without --read-only or --staged, a write tool (deletes included) runs as soon as the AI app's approval prompt is answered: the app's prompt replaces the CLI's own question. Do not set the app to always allow the delete tools. With --staged, you apply a proposal yourself with p202 change apply; the assistant cannot.

Claude Desktop and Cursor take the same command in their MCP settings: {"mcpServers": {"prosper202": {"command": "/path/to/p202", "args": ["mcp", "--read-only"]}}}

p202 mcp tools lists the tools; p202 mcp call <tool> --input <json> runs one.

06Commands

2 under p202 mcp.

p202 mcp callRun one MCP tool with JSON arguments, as an AI app's call wouldp202 mcp toolsList the MCP tools as JSON (names, input schemas, read/write hints)

07For agents

Running this from an agent

  • Read the same facts as JSON: p202 commands mcp --json.
  • With AI_AGENT, CLAUDECODE or another agent variable set, output is compact JSON and errors arrive on stderr as a JSON envelope with a hint.
  • Exit codes: 0 ok, 1 bad input, 2 auth, 3 network, 4 server error, 5 partial failure, 6 no Pro licence on the install.
p202 commands mcp --json
{
  "path": "p202 mcp",
  "use": "mcp",
  "short": "Serve Prosper202 to AI assistants over MCP (stdio), for Claude Desktop, Claude Code, Cursor and others",
  "long": "Runs an MCP (Model Context Protocol) server on stdin/stdout. An AI app starts it\nand calls its tools, each of which runs one p202 command against your install with\nthe profile that is active when the server starts (or --profile).\n\n  claude mcp add prosper202 -- p202 mcp --read-only    (recommended to start)\n  claude mcp add prosper202 -- p202 --staged mcp       (writes become proposals)\n  claude mcp add prosper202 -- p202 mcp\n\nWithout --read-only or --staged, a write tool (deletes included) runs as soon as\nthe AI app's approval prompt is answered: the app's prompt replaces the CLI's own\nquestion. Do not set the app to always allow the delete tools. With --staged, you\napply a proposal yourself with `p202 change apply`; the assistant cannot.\n\nClaude Desktop and Cursor take the same command in their MCP settings:\n  {\"mcpServers\": {\"prosper202\": {\"command\": \"/path/to/p202\", \"args\": [\"mcp\", \"--read-only\"]}}}\n\n`p202 mcp tools` lists the tools; `p202 mcp call <tool> --input <json>` runs one.",
  "runnable": true,
  "flags": [
    {
      "name": "read-only",
      "type": "bool",
      "default": "false",
      "usage": "Serve only the tools that read; writes are refused",
      "required": false
    }
  ]
}