Claude MCP for Codex
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| CLAUDE_BIN | No | Executable name or path, without extra arguments. | claude |
| CLAUDE_CONFIG_DIR | No | For an existing login stored in a nonstandard Claude configuration directory, forward your existing `CLAUDE_CONFIG_DIR` through Codex's `env_vars`; the wrapper does not inspect that directory. | |
| CLAUDE_TIMEOUT_MS | No | Request timeout; integer from 1,000 to 600,000 milliseconds. | 120000 |
| CLAUDE_DEFAULT_MODEL | No | Optional model alias or ID used when the tool omits `model`. | |
| CLAUDE_DEFAULT_EFFORT | No | Optional `low`, `medium`, `high`, `xhigh`, or `max` used when the tool omits `effort`. |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| ask_claudeA | Ask Claude Code for a second opinion using its existing local login. Send only sanitized text: never credentials, secrets, or customer data. Claude receives only the supplied prompt and cannot use tools. For reviews, return findings and recommendations; Codex handles code changes. Pass model/effort from applicable AGENTS.md guidance to override configured defaults. Successful calls include requested settings and CLI-reported model IDs; effective model and effort remain unverified. Each call starts a fresh conversation and consumes Claude usage. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 1 tool
With only one tool in the set, there is no possibility of overlap or misselection. ask_claude has a single, unambiguous purpose: obtaining a second opinion from Claude.
The lone tool name ask_claude follows a clean snake_case verb_noun convention that is self-descriptive. No competing conventions exist to create inconsistency.
A single tool is borderline thin by the rubric's standards, though the server's stated purpose (one-shot consultation with Claude) is narrow enough that one operation is defensible. There is no room to over- or under-scope further.
For the declared purpose of a stateless second-opinion bridge, the surface is effectively complete: the one operation exists and its limits are documented. Minor gaps exist around multi-turn follow-up or session continuity, but those appear intentionally excluded and agents can work around them.