Description
Hi, thanks for open-sourcing MetaBot — this project looks very interesting.
From the README and setup flow, it seems MetaBot is currently deeply integrated with Claude Code / Claude Code CLI as the agent runtime. I’m wondering whether you would consider supporting OpenAI Codex as an alternative backend as well. (GitHub)
Use case
I’d like to use MetaBot’s infrastructure layer — such as:
• IM bridge
• shared memory
• agent factory / skill system
• scheduler
• agent communication bus
but with Codex as the coding agent/runtime in some environments.
Possible reasons:
• some teams already standardize on OpenAI tooling
• model/provider flexibility would make MetaBot easier to adopt
• it would be useful to compare Claude Code and Codex under the same orchestration framework
Questions
1. Is Codex support already possible through some adapter or custom configuration?
2. If not, would you be open to supporting it in the future?
3. Architecturally, is the current agent runtime abstract enough to support multiple backends, or is it tightly coupled to Claude Code today?
4. If you would accept a PR for this, what integration boundary would you recommend?
Possible direction
Maybe the runtime layer could be abstracted so MetaBot can plug into different coding agents, for example:
• Claude Code
• OpenAI Codex
• other compatible agent CLIs
Additional context
I’m not asking for Claude Code to be replaced — only whether a pluggable backend / multi-provider runtime could be supported.
Thanks!
Description
Hi, thanks for open-sourcing MetaBot — this project looks very interesting.
From the README and setup flow, it seems MetaBot is currently deeply integrated with Claude Code / Claude Code CLI as the agent runtime. I’m wondering whether you would consider supporting OpenAI Codex as an alternative backend as well. (GitHub)
Use case
I’d like to use MetaBot’s infrastructure layer — such as:
• IM bridge
• shared memory
• agent factory / skill system
• scheduler
• agent communication bus
but with Codex as the coding agent/runtime in some environments.
Possible reasons:
• some teams already standardize on OpenAI tooling
• model/provider flexibility would make MetaBot easier to adopt
• it would be useful to compare Claude Code and Codex under the same orchestration framework
Questions
Possible direction
Maybe the runtime layer could be abstracted so MetaBot can plug into different coding agents, for example:
• Claude Code
• OpenAI Codex
• other compatible agent CLIs
Additional context
I’m not asking for Claude Code to be replaced — only whether a pluggable backend / multi-provider runtime could be supported.
Thanks!