OpenAI Codex CLI

OpenAI Codex CLI is OpenAI's command-line coding assistant. It supports MCP servers through global configuration.

Configuration

Codex CLI uses global configuration only - there is no project-level MCP config support.

Scope	Location	Format
Global	~/.codex/config.toml	TOML

Configure the MCP server path based on your platform:

macOS

[mcp_servers.sailfishTelemetry]
command = "node"
args = ["/Applications/SF Veritas.app/Contents/Resources/mcp-server/index.js"]
enabled = true

Windows

[mcp_servers.sailfishTelemetry]
command = "node"
args = ["%LOCALAPPDATA%\\Programs\\SF Veritas\\resources\\mcp-server\\index.js"]
enabled = true

Linux

[mcp_servers.sailfishTelemetry]
command = "node"
args = ["/opt/SF Veritas/resources/mcp-server/index.js"]
enabled = true

Verification

After configuring, verify the connection:

1. Open your terminal
2. Run codex to start the CLI
3. Type /mcp to see connected MCP servers
4. You should see sailfishTelemetry listed

Usage

Once connected, ask Codex to use the telemetry tools:

> List my telemetry sessions
> Show me errors from the latest session
> What functions are taking the longest to execute?
> Analyze network requests to external APIs

Codex will use the MCP tools to query your local telemetry data.

Troubleshooting

Server not appearing in /mcp

1. Verify the app is installed at the expected location
2. Check TOML syntax - ensure quotes around string values
3. Verify enabled = true is set
4. Restart Codex CLI

Config file not found

Create the config directory and file:

mkdir -p ~/.codex
touch ~/.codex/config.toml

"No telemetry data found"

1. Ensure your application has been instrumented with SF Veritas SDK
2. Run your application to generate telemetry data
3. Check that the database exists: ls ~/.sailfish/telemetry.db