all-agents-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| AA_MCP_LOG_LEVEL | No | Log level (debug, info, warn, error) | info |
| AA_MCP_CODEX_MODELS | No | Available Codex models | gpt-5.3-codex-spark,gpt-5.3-codex,gpt-5.2-codex-max,gpt-5.2-codex |
| AA_MCP_CLAUDE_MODELS | No | Available Claude models (comma-separated) | claude-opus-4.6,claude-sonnet-4.5,claude-haiku-4.5 |
| AA_MCP_CODEX_DEFAULT | No | Codex default model | gpt-5.3-codex-spark |
| AA_MCP_GEMINI_MODELS | No | Available Gemini models | gemini-3-pro-preview,gemini-3-flash-preview,gemini-2.5-pro,gemini-2.5-flash |
| AA_MCP_CLAUDE_DEFAULT | No | Claude default model | claude-opus-4.6 |
| AA_MCP_COPILOT_MODELS | No | Available Copilot models | claude-opus-4.5,claude-sonnet-4.5,claude-haiku-4.5,gpt-5.2-codex,gemini-3-pro-preview,gemini-3-flash-preview |
| AA_MCP_GEMINI_DEFAULT | No | Gemini default model | gemini-3-pro-preview |
| AA_MCP_COPILOT_DEFAULT | No | Copilot default model | claude-sonnet-4.5 |
| AA_MCP_CODEX_ANALYSIS_LEVEL | No | Codex reasoning depth (low/medium/high/xhigh) | xhigh |
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
} |
| resources | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| ask_agentA | Ask a specific AI agent a question. Use when targeting a single agent like 'ask codex' or 'ask gemini'. |
| ask_allA | Ask multiple agents the same question in parallel. Use for explicit multi-agent comparison. |
| delegate_taskA | Delegate a task to agent(s). Automatically analyzes complexity and routes to single or parallel execution. |
| collaborateC | Collaborate with an agent for analysis. Sends prompt, gets response, and provides guidance for comparison. |
| verifyA | Cross-verify by running the same prompt across multiple models of one agent. Compares results for consistency. |
| review_codeC | Review code using a specific agent. Analyzes for bugs, security, performance, and clarity. |
| debug_withB | Debug an error using a specific agent. Provide the error and optional code context. |
| explain_withB | Get code explanation from a specific agent. Choose brief or detailed level. |
| fetch_pageA | Fetch web page content via Gemini CLI. Gemini has native web browsing — use this to retrieve and extract content from URLs. |
| generate_testA | Generate tests for code using a specific agent. Auto-detects or specify test framework. |
| refactor_withC | Refactor code using a specific agent. Optionally specify a goal: performance, readability, or modularity. |
| list_agentsA | List all detected agents and their availability status. |
| list_modelsB | List available models for agents. Optionally filter by specific agent. |
| agent_healthA | Check agent health status including availability, authentication, and latency. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
| sessions-list | |
| agent-status |
TDQS
Scored across 14 tools
The 'ask' family of tools (ask_all, ask_agent, delegate_task, collaborate, verify) has overlapping purposes, especially between ask_all/delegate_task (both parallel) and ask_agent/delegate_task (single vs auto-routing). Code and meta tools are clearly distinct, but the query tools could cause misselection without careful reading.
Names mix patterns: verb_noun (ask_agent, review_code), verb_with (debug_with, explain_with, refactor_with), standalone verbs (collaborate, verify), and noun-based (agent_health). While most follow a verb-leading style, the suffixes and noun-only names break predictability.
14 tools is near the upper bound but reasonable for a multi-purpose agent integration server that covers querying, code tasks, web fetch, and system info. Each tool has a role, though some redundancy in the 'ask' cluster makes the count feel slightly high.
The server covers the main workflows: asking agents, delegating, collaborating, verifying, code review/debug/explain/test/refactor, web fetch, and agent/model discovery. Minor gaps exist (e.g., no task cancellation or history retrieval) but they are not critical for core usage.