Codex ACP Agent
Supports UIby cola-io
Bridge between OpenAI Codex runtime and ACP clients via stdio for advanced agent orchestration.
What it does
Connects the OpenAI Codex runtime with Agent Client Protocol (ACP) clients, allowing AI agents to operate within IDEs and other ACP-compatible environments. It manages session state, handles slash commands, and provides a unified interface for agent-client communication.
Tools
read_text_file: Reads workspace files via ACP or local disk.write_text_file: Writes workspace files via ACP or local disk.edit_text_file: Performs focused text replacements in files.multi_edit_text_file: Applies multiple sequential replacements to files.
Installation
Add to Zed settings:
"agent_servers": {
"Codex": {
"command": "/path/to/codex-acp",
"args": [],
"env": {
"RUST_LOG": "info"
}
}
}
Supported hosts
- Zed
- Codex
Choose your AI client and follow the steps below.
zed
Add the agent_servers configuration for Codex to your zed settings.jsonRelated Apps
FinanceToolkit
MCP ServerProfessional-grade financial analysis toolkit for equities, options, and risk management.
DiffSitter MCP
MCP ServerAI-powered structural code navigation using tree-sitter ASTs for semantic understanding across 14+ languages.
OpenAI Apps SDK Examples
MCP AppOfficial example gallery of interactive MCP widgets for ChatGPT — 3D viewers, maps, carousels, shopping carts, and more.
Human MCP
MCP ServerGive AI agents human-like senses: visual analysis, image/video generation, speech synthesis, browser automation, and advanced reasoning — 29 MCP tools in one se
Shopify MCP Server
MCP ServerDirect interaction with Shopify store data via GraphQL API for managing products, customers, and orders.
Git MCP Server
MCP ServerFull-featured Git MCP server exposing 28 tools for AI agents to clone, commit, branch, diff, merge, rebase, and more via STDIO or Streamable HTTP.
CodexPotter
MCP ServerAutonomous reconciliation loop that drives Codex to align your codebase with instructed states.
Smart Coding MCP
MCP ServerSemantic code search for AI assistants using local embeddings — find code by meaning, not just keywords.