Using with Codex CLI

The easy way

copillm codex

That handles login check, daemon start, agent install/resolve, and env wiring.

Manual wiring

If you’d rather wire it up yourself, copillm env codex prints the env block:

$ copillm env codex
# Codex CLI → copillm
export CODEX_HOME="/Users/you/.copillm/codex"

copillm start already generates ~/.copillm/codex/config.toml with the right [model_providers] block for live discovery against the local proxy.

--shell fish and --shell powershell are also supported, and --json returns a machine-readable payload.

Generated config.toml

The auto-generated ~/.copillm/codex/config.toml points Codex’s model provider at http://127.0.0.1:4141/codex/v1 and enables live model discovery via the /codex/v1/models endpoint. You can override CODEX_HOME to point Codex at a different config location if you want to manage it yourself.

To make the default Codex install use copillm without launching through copillm codex, run:

copillm config sync --agent codex

That merges the generated copillm provider block into ~/.codex/config.toml and applies the active profile’s MCP servers there.

Pass-through args

copillm codex --model gpt-5
copillm codex --help

Any extra flags or positional args are forwarded verbatim to the codex binary.