x0c

Agent Session Control — Development skill for Claude Code

Development community

Python launch plans, audited runs and ACP/MCP wiring for Claude, Codex, OpenCode, Cursor, Pi.

How to install Agent Session Control

This entry records only its repository, not the path inside it, so there is no exact command to give. Open x0c/agent-session-control and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Agent Session Control does

Python launch plans, audited runs and ACP/MCP wiring for Claude, Codex, OpenCode, Cursor, Pi.

Alternatives in Development

Agentic Stack — One brain, many harnesses 2.2k ★
Zeron — A native control plane for Claude Code, Codex, Cursor, Devin and other coding agents 1.4k ★
OpenSquirrel — For people who get distracted by agents 1.3k ★

README

Agent Session Control

**Languages:** English | [简体中文](README.zh-CN.md)

Launch and track local **Claude Code**, **Codex CLI**, **OpenCode**, **Cursor**, and **Pi** coding-agent sessions from Python: pure launch plans, audited runs, and per-session stdio MCP wiring. Also ships an [agent-client-protocol](https://github.com/agent-client-protocol/agent-client-protocol) (ACP) adapter.

This is a **Python library + JSON CLI + ACP adapter** — not a TUI, not a session manager, not a history parser (history observation is delegated to [SessKit](https://github.com/x0c/sesskit)).

[![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)

Status: 0.1.0 released and tested (see [docs/DESIGN.md](docs/DESIGN.md) and [docs/CONTRACT.md](docs/CONTRACT.md)). This README claims only what the release delivers and its tests prove; per-runtime limits are listed below.

What works today (tested)

Pure launch plans (agent_session_control.runtime): build_create_plan and build_resume_plan for the five built-in runtimes (claude, codex, opencode, cursor, pi) plus explicit third-party RuntimeProfile registration. Plans are immutable argv arrays — building one never launches a process, mutates storage, edits global config, or sends a model request. Caller defaults are preserved (no model flag, no approval relaxation unless requested). Unsupported options — unknown runtimes, --model on OpenCode, preassigned native ids where no flag exists — fail before launch instead of silently degrading.
Per-session stdio MCP (mcp_servers=): verified launch-only channels, each proven end-to-end with a disposable stdio fixture against the real installed CLI (no global config touched). Secrets never travel on argv — Claude argv carries ${NAME} references, Codex argv carries env_vars forwarding names, OpenCode carries the document in private process env; live values travel only in plan.env, redacted from every display proje