A feature-rich, well-documented community MCP server for DeepSeek with dual V4 models, session management, schema-validated JSON, cost tracking, and circuit-breaker fallback, backed by ~89% test coverage. Ideal for developers integrating DeepSeek into MCP workflows, though note it is unofficial and self-hosted HTTP mode requires careful endpoint security.
Strongest · Documentation 17/20 Weakest · Reliability 9/20
Static review · not runListed 2026-09-07
Read the FMRS scoring method →Runtime：Node.js 20+ · Docker · Hosted (no local runtime)
DEEPSEEK_API_KEY
DeepSeek API key, required; get one at https://platform.deepseek.com.
HTTP_AUTH_TOKEN
Static gateway token for the self-hosted HTTP endpoint; when set, POST /mcp requires Bearer auth.
DEEPSEEK_BASE_URL
Custom DeepSeek API endpoint, defaults to https://api.deepseek.com.
DEFAULT_MODEL
Default model for requests, defaults to deepseek-v4-flash.
SHOW_COST_INFO
Whether to show cost info in responses, default true.
REQUEST_TIMEOUT
Request timeout in milliseconds, default 60000.
MAX_RETRIES
Maximum retries for failed requests, default 2.
SKIP_CONNECTION_TEST
Skip the startup API connection test, default false.
MAX_MESSAGE_LENGTH
Maximum message content length in characters, default 100000.
SESSION_TTL_MINUTES
Session time-to-live in minutes, default 30.
MAX_SESSIONS
Maximum number of concurrent sessions, default 100.
FALLBACK_ENABLED
Enable automatic model fallback on errors, default true.
CIRCUIT_BREAKER_THRESHOLD
Consecutive failures before the circuit opens, default 5.
CIRCUIT_BREAKER_RESET_TIMEOUT
Milliseconds before the circuit half-opens, default 30000.
MAX_SESSION_MESSAGES
Max messages per session (sliding window), default 200.
RESPONSE_SCHEMA_MAX_RETRIES
Repair retries when response_schema validation fails, default 2, 0 disables.
ENABLE_MULTIMODAL
Enable multimodal (image) input support, default false.
TRANSPORT
Transport mode: stdio or http, default stdio.
HTTP_PORT
HTTP server port when TRANSPORT=http, default 3000.
HTTP_HOST
Bind address for HTTP transport, default 127.0.0.1.
HTTP_ALLOWED_HOSTS
Comma-separated allowed Host headers for DNS rebinding protection when binding to 0.0.0.0.
Requires Node.js 20+ and a DeepSeek API key (from platform.deepseek.com). Claude Code local: claude mcp add -s user deepseek npx @arikusi/deepseek-mcp-server -e DEEPSEEK_API_KEY=your-key-here. Gemini CLI: gemini mcp add deepseek npx @arikusi/deepseek-mcp-server -e DEEPSEEK_API_KEY=your-key-here. Or use the hosted endpoint (no install): claude mcp add --transport http deepseek https://deepseek-mcp.tahirl.com/mcp --header "Authorization: Bearer YOUR_DEEPSEEK_API_KEY".
{"mcpServers":{"deepseek":{"command":"npx","args":["@arikusi/deepseek-mcp-server"],"env":{"DEEPSEEK_API_KEY":"your-api-key-here"}}}}
Shown for Claude Code. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.
{
"servers": {
"deepseek": {
"command": "npx",
"args": [
"@arikusi/deepseek-mcp-server"
],
"env": {
"DEEPSEEK_API_KEY": "your-api-key-here"
}
}
}
}
Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).
claude mcp add deepseek -e DEEPSEEK_API_KEY=your-api-key-here -- npx @arikusi/deepseek-mcp-server
Run it in a terminal; replace any <…> placeholders with your own values first.
After install, the client's tool list should show deepseek_chat, deepseek_fim, and deepseek_sessions; asking a question answered by DeepSeek with token and cost info in the response confirms the connection works.
Once connected, you can ask your AI assistant things like:
No matching tools
Listed from the project's documentation, not tested by this site.
DeepSeek MCP Server is a community-maintained, MIT-licensed MCP server for accessing DeepSeek V4 models (deepseek-v4-flash and deepseek-v4-pro), both with a 1M-token context window and optional chain-of-thought thinking mode. It provides multi-turn sessions, function calling (up to 128 tool definitions), JSON output with schema validation, fill-in-the-middle (FIM) completion, cache-aware cost tracking, automatic model fallback with circuit breaker protection, 12 prompt templates, and 3 MCP resources (models, config, usage). It runs via npm (stdio) or a hosted BYOK endpoint (Streamable HTTP), and works with Claude Code, Gemini CLI, Cursor, Windsurf, and other MCP clients. Note: this is an independent community project, not an official DeepSeek product.
Upstash's official server providing up-to-date third-party library docs for AI coding assistants
AI-powered NPM package analysis MCP server
An independent control layer that gives AI coding agents budgets, brakes, and receipts
GitHub's official MCP server for managing repos, issues, PRs, and workflows via natural language
Source revision 1bac160857d5 Data synced 2026-10-19 Read the FMRS scoring method