Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@dsh-codex-bridge start a code review task on the current branch"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
DSH 插件 / MCP 服务器 — 让 Codex、Claude Code、Cursor 等 MCP 客户端将 DSH 当作外部可恢复子代理使用。
flowchart LR
C[Codex / Claude Code / Cursor / any MCP client] -->|MCP SSE| B[DSH Bridge :3101]
B -->|spawn detached| S[Supervisor]
S -->|dsh --profile headless| W[DSH Worker]
S -->|持久化| F[(~/.dsh/tasks/)]5 MCP tools: dsh_task_start, dsh_task_poll, dsh_task_cancel, dsh_task_list, dsh_get_status
Async task model: start returns immediately, poll for events via cursor
Resumable: bridge restart recovers running tasks, orphans dead ones
Visual dashboard: http://127.0.0.1:3101/agents — frosted glass UI
REST API: full task management API
SSE stream: real-time task events
Security: bearer token, 127.0.0.1 only, cwd restriction
Related MCP server: iStdBMAD
# 1. Install as DSH plugin
dsh plugin --profile web add dsh-codex-bridge
# 2. Set token (optional, has dev default)
export DSH_CODEX_BRIDGE_TOKEN="your-32-char-token"
# 3. Start DSH
dsh web
# 4. Open dashboard
open http://127.0.0.1:3101/agentsTool | Description |
| Start an async task, returns taskId + dashboardUrl |
| Read events incrementally by cursor (≤64 KiB) |
| Cancel a running task |
| List all tasks with status |
| Bridge health and stats |
queued → running → succeeded
→ failed
→ timed_out
→ cancelled
→ orphaned[mcp_servers.dsh_agent]
url = "http://127.0.0.1:3101/sse"
bearer_token_env_var = "DSH_CODEX_BRIDGE_TOKEN"
startup_timeout_sec = 10
tool_timeout_sec = 45
enabled_tools = [
"dsh_task_start",
"dsh_task_poll",
"dsh_task_cancel",
"dsh_task_list",
"dsh_get_status"
]Method | Path | Description |
GET |
| List tasks |
GET |
| Task detail |
POST |
| Create task |
POST |
| Cancel task |
POST |
| Retry task |
GET |
| Get events (cursor) |
GET |
| Download artifact |
GET |
| SSE event stream |
The dashboard at http://127.0.0.1:3101/agents provides:
Left panel: task list with status filter
Center: real-time event stream (stdout/stderr/status)
Right panel: task details (PID, timing, artifacts)
Dark/light mode: follows system preference
dsh-codex-bridge/
├── lib/
│ ├── index.js DSH plugin entry + MCP server + REST API
│ ├── dashboard.html Visual dashboard (single-file HTML/CSS/JS)
│ └── supervisor.mjs Detached supervisor process
├── package.json
├── cordis.patch.yml
└── LICENSEMIT
This server cannot be deployed
Official MCP server for Agentwork — delegate tasks to AI agents with human-in-the-loop
Nifty's MCP server — exposes tasks, projects, messages, and files as tools for AI agents.
Hosted MCP server connecting claude.ai, ChatGPT and other AI apps to your own computer
Hosted MCP server connecting AI assistants to 9,000+ apps and 40,000+ actions via Zapier.
