t3code-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| T3_MCP_STATE_DIR | No | Selects a different directory for the bridge's state files (credentials, proof key, operation journal). Defaults to ~/.t3code-mcp. | ~/.t3code-mcp |
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 |
|---|---|
| list_environmentsA | List machines linked to the signed-in T3 account. Availability probes are optional and bounded to this page. |
| list_projectsA | List projects and their model defaults on one explicitly selected environment. |
| list_threadsA | Discover unarchived threads on one environment with IDs, titles, current status, approval/input flags and recent turn timing. Optionally filter by project or status. Newest activity first; bounded pages. Use get_thread with a returned threadId to read messages. Finished means the latest turn completed, not verified task success. |
| start_threadA | Create a thread and send instructions to its agent in the existing project workspace. May change files and incur provider charges. Defaults to approval-required. Worktree creation is not supported. Reuse operationId and identical arguments after ambiguous failures. |
| get_threadA | Read bounded recent messages, activity, current status, provider errors and requests for human attention. Discover thread IDs with list_threads. Older turns can be paginated using beforeCursor. Reply to approvals and input in T3. |
| send_messageA | Send follow-up instructions using the thread's existing model and runtime modes. May change files or incur charges. Preserve operationId and arguments for retries. |
| interrupt_threadA | Request interruption of the current or explicitly selected turn. The target turn is recorded so a retry cannot interrupt a later turn. |
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 7 tools
Most tools are clearly distinct: list_threads/get_thread for reading, start_thread/send_message for creating and continuing threads, interrupt_thread for control, and list_environments/list_projects for discovery. The only mild overlap is between list_threads and get_thread, but their descriptions make the list-vs-detail relationship clear.
Tool names follow a consistent verb_noun pattern: list_threads, get_thread, interrupt_thread, list_environments, list_projects, send_message, start_thread. The pattern is predictable, though 'send_message' and 'start_thread' are slightly less uniform than the list/get/interrupt group.
Seven tools is well-scoped for a T3 coding-agent MCP server. Each tool covers a distinct operation: discovery, reading, starting, sending, and interrupting threads, plus environment and project listing. No tool feels redundant or unnecessary.
The surface covers the core thread lifecycle: list, get, start, send, and interrupt. Minor gaps exist, such as no explicit archive/unarchive or delete/cancel thread operation, and worktree creation is noted as unsupported, but agents can accomplish the main workflows.