agentmux
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| AGENTMUX_HOME | No | Override the directory used to store local session metadata (default is ~/.agentmux). |
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 |
|---|---|
| whoamiA | Return the managed agent identity inherited by this MCP process, or managed=false for an external control-tower host. |
| spawnC | Spawn a coding-agent session and start its first job asynchronously. Managed callers automatically create children in their own team. |
| spawn_manyC | Spawn up to 16 coding-agent sessions. Sessions are started sequentially but their jobs run concurrently. |
| sendC | Continue an existing idle provider-native session. Managed callers are limited to their team. |
| message_sendB | Persist an attributed message to another agent. With wake=true, also start a new turn if the recipient is idle and resumable. |
| inboxA | Read persisted messages. Managed agents can only read their own inbox; external supervisors must specify agent_id. |
| message_ackB | Mark one or more inbox messages as read. |
| waitA | Wait for one or more jobs to finish, returning early when all are terminal. Timeout is capped at 60 seconds. |
| statusB | Get one accessible agent and its latest job. |
| resultB | Get an accessible job result by job ID. |
| listA | List visible sessions. Managed agents see only their team; external supervisors see all sessions. |
| killA | Cancel an active job and stop a session. Managed callers may stop only themselves or their descendants. |
| team_createC | Create a top-level logical team. This tool is available to external control-tower hosts. |
| team_statusA | Get a team and its members. Managed agents are limited to their own team. |
| team_listA | List visible teams. Managed agents see their own team; external supervisors see all teams. |
| providersA | List provider adapters and the local CLI command each adapter expects. |
| doctorA | Check whether each supported provider CLI is installed and report its version. |
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 17 tools
Most tools target distinct resources and actions: sessions, jobs, teams, messages, and provider metadata are clearly separated. The main ambiguity is between send and message_send, but their descriptions clarify session continuation versus persisted messaging.
Naming is readable but mixes conventions: verb_noun (team_create, team_status), noun_verb (message_send, message_ack), bare verbs (send, wait, list, kill), and nouns (inbox, status, providers). The send/message_send pair is especially inconsistent.
With 17 tools, the server is at the borderline of feeling heavy for an orchestration/mux utility. Each tool covers a distinct concern, but the count is above the typical well-scoped range.
The surface covers core session, job, team, and messaging workflows, but there are notable gaps: no way to list all jobs, no team deletion or update, and no session transcript/history retrieval. These gaps can force workarounds but basic orchestration is functional.