"Information on agent swarm architecture" matching MCP servers:

Bridges the ChatGPT web UI to a local development environment over MCP, letting the model read and patch files, run real shell commands and terminals, execute tests, and control the desktop (screen, mouse, keyboard, clipboard) within user-approved folders. It adds parallel workers, goal/loop long-task control, plugin tools, and a security-hardening layer with risk-tiered shell policy, audit logging, and credential sanitization.

Connects ChatGPT to local files, terminals, and the desktop through MCP with Chrome integration, letting users read and edit project files, run commands, and follow tool results as they arrive. It also adds durable multi-agent workers, Goal/Loop task tracking, mid-task corrections, and Compact & Resume for carrying session and worker history into a fresh chat.

Enables ChatGPT's web interface to operate a real Mac by executing shell commands and reading or writing files over a reverse SSH tunnel, without installing a client or consuming API quota.

Enables code execution in isolated Docker containers with persistent IPython, Node.js, or R kernels, supporting file import/export and cross-session transfers via MCP tools.

Provides workspace-scoped file operations and command execution tools for building Cursor-style agents, along with system prompts and tool definitions for LLM integration.

Delegates coding and read-only exploration tasks to the Antigravity CLI as blocking headless runs, mirroring live progress to on-disk stream transcripts and diffing git status to report touched files and out-of-scope edits. Supports resumable conversations, per-call model/agent/effort overrides, idle and hard timeouts, and discovery of available models and agent profiles.

Encoding-aware, indentation-smart file tools for AI coding agents. 20+ tools including read/edit with automatic encoding detection, smart indentation conversion, SSH, SFTP, process management, and system utilities.

Enables MCP-based communication with existing interactive terminal channels, such as shells or agent CLIs, by listing, reading, writing text, and sending control signals to tmux sessions without creating or managing them.

Enables MCP clients to drive developer workflows through APIs and CLIs, with explicit workspace or trusted-workstation launch modes, bounded command previews, multi-repository rules discovery, and an on-demand skills/recipes entry point. It lets agents review changes, run commands, and load only the smallest relevant rule or skill while reporting evidence and unverified steps.

Discovery router for the agent-device CLI with status, install, help, prompts, and resources.

A download tool for AI agents that wraps aria2 with safety rules and MCP server support, enabling reliable downloads without false positives.

Exposes tools over stdio that let an AI agent create disposable Docker sandboxes to run commands and stream output, read and write files, and take snapshots, roll them back, fork them, list, and destroy them. It tells the model to snapshot before risky changes so failed attempts can be undone or explored in parallel on self-hosted infrastructure.

Creates and manages multi-agent teams of AI coding assistants (Claude Code, Codex CLI) that collaborate in tmux terminal sessions. Enables a leader agent to assign sub-tasks to team members for coordinated development.

Enables Notion AI to read and modify files in a single local PC folder and, in full mode, run commands and Codex, Claude Code, or Gemini CLI agents, with user confirmation for each action.

Enables coding agents and users to share live tmux-based REPL sessions for many languages, allowing agents to execute code in the same interactive workspace via MCP.

Local automation bridge that is both an MCP server and client: 234 tools across 44 namespaces for shell, filesystem, git, desktop control, browser (CDP), Android (ADB), OCR/ASR and memory, plus the ability to connect external MCP servers. Sandboxed execution with token auth, fail-closed policy and audit log.

Local daemon that bridges AI assistants to the local filesystem with read-only commands and optional write mode, enforcing strict security filters to prevent credential leakage.

Gives ChatGPT hands-on control of a local machine: reading, patching and executing code in approved project folders, running real interactive processes, managing plans and worker sub-chats, and driving the browser or desktop through user-granted permissions. Every capability is a switch inside an approved-folder sandbox, so nothing runs that the user did not explicitly enable.

MCP server for persistent SSH sessions with tmux integration, port forwarding, and SFTP file writes, built for AI agent workflows.

An MCP server that bridges CLI coding agents like Claude Code, Codex, opencode, and Antigravity into any MCP client, enabling synchronous and asynchronous task execution, follow-up input, and a structured code review tool.

MCP server for controlling iTerm2 terminal sessions, enabling screen reading, command execution, keystroke sending, and session management through natural language.

Enables AI models to operate on a Windows PC through local filesystem and terminal MCP tools, with token auth and active project switching.

MCP wrapper for the agent-msg.sh file-based message bus, enabling MCP-compatible clients to send, receive, and manage inter-agent messages with bound identity and HMAC-signed payloads.

A programmable terminal control plane that enables AI agents to orchestrate, monitor, and interact with multiple parallel AI CLI sessions and browser instances within CMUX. It provides over 80 tools for workspace management, pane manipulation, and cross-agent communication to facilitate complex multi-project workflows.

Enables any MCP client to execute DeepSeek Harness agents and shell commands through standard MCP tools, bridging multi-agent capabilities with sandboxed execution.

Enables ChatGPT web to use local tools like file reading, command execution, and patch application through an MCP server over OpenAI Secure MCP Tunnel.

Enables AI agents to safely execute privileged commands by routing them to a human approver, who runs them outside the agent's context and returns the output to the model.

An MCP server that exposes a sandboxed task runner protocol to any MCP client, enabling evidence-protected task execution, exit protocol reading, and autopsy reports.

Enables ChatGPT web to interact with local Windows/WSL shell and code workspaces via an MCP server, providing file access, shell execution, and snapshot-based workspace management with per-command authorization.

An MCP-only server that launches AI CLI tasks (Claude, Codex, Gemini, Forge, OpenCode) as background subprocesses and manages them via PID for status, output, and lifecycle control.