Coding agents · MCP server
Deepseek as subagent: MCP server, setup and security
Deepseek as subagent is an open-source MCP server published on GitHub that lets an AI agent delegate file reading and editing to DeepSeek in a sandboxed workspace. Its repository has 51 stars and its last commit is from October 22, 2026 and it is licensed under MIT.
| Adoption (stars) | 17/40 | |
|---|---|---|
| Maintenance | 30/30 | |
| Framework (licence, registry, official) | 15/30 |
What Deepseek as subagent does
According to its repository: “Run DeepSeek as a sub-agent in Claude Code or Codex CLI, with its own agent loop for reading, editing and running commands in a sandboxed workspace.”
The Model Context Protocol (MCP) is a standard plug between an AI agent and an outside service: once the server is connected, its tools show up in the agent, which decides when to call them. Deepseek as subagent is listed under Coding agents, a category of 151 servers in the directory, where it ranks #107 by stars.
Topics declared by the repository: agent-loop, ai-agent, claude-code, codex-cli, deepseek, llm-orchestration.
Installing and connecting Deepseek as subagent
No package or remote server is declared in the official MCP registry: you install it from source, following the repository README.
It works with any MCP-compatible client (Claude Desktop, Claude Code, Cursor, VS Code, Codex and most recent agents). You add one entry to the client's MCP configuration: server name, start command or URL, environment variables. Project language: Python. Listed systems: macOS, Windows, Linux.
Security: what to check
Connecting an MCP server means installing software that talks straight to your agent: it runs with your rights and the model trusts its tool descriptions. The risk therefore depends on what it can touch and how far its code can be trusted.
None of the risk signals we track (system access, secrets, missing licence, abandonment) applies to Deepseek as subagent. That does not replace reading the code.
Good practice: grant only the rights needed (least privilege), require human approval for irreversible actions, and run the AI agent security check.
Project activity and maintenance
Created on May 11, 2026, the PsChina/deepseek-as-subagent repository has 51 stars and 14 forks, with 0 open issues. It has more stars than 52% of the servers in the directory. It received changes this month.
Stars measure attention, not quality: a recent, heavily promoted project can collect them fast, while a useful but niche server will have few.
Verdict
Deepseek as subagent is a reasonable choice; check the points below before connecting it.
Strengths
- Active project: recent commit
- Permissive licence (MIT)
Limitations
- Install from source
Alternatives
- Desktop Commander10.1K ★
- Ouroboros6.3K ★
- Repomix29K ★
- Claude Code Ultimate Guide6.2K ★
- Harness1.4K ★
- Appllama3K ★
Learn more
Frequently asked questions
How do I install Deepseek as subagent?
No package is declared in the MCP registry: clone the PsChina/deepseek-as-subagent repository, follow its README, then declare the start command in your client's MCP configuration.
Is Deepseek as subagent free?
The code is published under the MIT licence and can be used free of charge.
Is Deepseek as subagent safe?
No risk signal found (system access, secrets, licence, abandonment). As with any MCP server, keep rights minimal and read the code before sensitive use.
Is it an official server?
Nothing says so: it is a project published by PsChina on GitHub. Check who maintains it before giving it access.
What are the alternatives to Deepseek as subagent?
In the same category: Desktop Commander, Ouroboros, Repomix, and Claude Code Ultimate Guide. Desktop Commander and Ouroboros score higher on our grid.
Related articles
- Build your own MCP server in 30 minutesOctober 24, 2026
- Automate your inbox with an AI agentOctober 18, 2026
- MCP, A2A, and tool calling: agent protocols explainedOctober 17, 2026
- Build an AI agent with n8n, without codingOctober 16, 2026
Neutrality. No affiliate links. No vendor pays to be reviewed, rated or ranked. Reviews are directed by François Chollet, neutral and factual, and every data point is dated and sourced. How we score