Skip to content
MCP server

Proxima

Local MCP server that routes ChatGPT, Claude, Gemini and local models into your editor

Pricing
Freemium
Platforms
Web, macOS, Windows
GitHub stars
1.2K
Last commit
1 month ago
Language
JavaScript
Listing updated
today

Similar MCP servers

About Proxima

Proxima is a freemium MCP server for ai models and llm tooling. It is available on Web, macOS, Windows, Linux, vscode, CLI, API and self-hosted.

The README says Proxima runs on 127.0.0.1 as a single gateway for several AI providers. Editors such as Cursor, VS Code, Windsurf and Claude Desktop connect to it over MCP, and it can reach models through logged-in browser sessions, through the user's own API keys, or through local runtimes like Ollama and LM Studio.

Beyond routing, it exposes 41 MCP tools for code review, test writing, error fixing, security audits and cross-model comparison, plus multi-model tools such as ask_all_ais and debate. A REST endpoint, WebSocket stream and terminal CLI use the same backend, and an optional Python agent handles multi-step coding and browser tasks with approval gates.

It suits developers who already pay for several AI chat subscriptions and want them available inside their coding tools, or who want to keep code on their own hardware. Session mode automates web logins, which the README admits may fall outside some providers' terms, so commercial teams are pointed toward BYOK.

Tools this server provides (25)

Taken from the project documentation; the running server may differ.

list_available_models
Discover real-time models across active providers with capability tags.
ask_chatgpt
Send a message to ChatGPT.
ask_claude
Send a message to Claude.
ask_gemini
Send a message to Gemini with thinking level control.
ask_perplexity
Send a message to Perplexity.
ask_model
Send to any session or BYOK provider.
ask_all_ais
Query all enabled providers simultaneously.
smart_query
Auto-route with verify, consensus or collaborate modes.
new_conversation
Reset a provider's conversation context.
generate_code
Generate code from a description.
review_code
Review a snippet for bugs, security and best practices.
explain_code
Line-by-line code explanation.
optimize_code
Performance optimization suggestions.
fix_error
Fix an error with code patches.
explain_error
Plain-English error explanation with fixes.
convert_code
Translate code between languages or frameworks.
write_tests
Generate test files.
security_audit
Security vulnerability audit.
deep_search
Typed search across web, reddit, github, news, academic and factcheck.
ddg_search
DuckDuckGo link search.
web_scrape
Convert a URL to Markdown with SSRF guarding.
run_workflow
Execute a multi-step workflow.
crew
Multi-AI collaborative task execution.
debate
Multi-perspective debate between models.
verify
Cross-AI fact-checking with confidence ratings.

What it does

Local MCP gateway
Runs a localhost MCP server that editors can call over stdio, with the core app providing the model connections.
Browser session routing
Uses existing logged-in ChatGPT, Claude, Gemini and Perplexity tabs so no API key is required in session mode.
Bring your own keys
Connects OpenAI, Anthropic, Google, DeepSeek, Groq, OpenRouter and other providers, with keys stored in the OS keychain.
Local model support
Routes requests to Ollama, LM Studio and other OpenAI-compatible local endpoints.
Multi-model tools
Sends one prompt to several providers, cross-checks answers or runs a debate between models.
REST, WebSocket and CLI access
Offers an OpenAI-compatible endpoint on port 3210, streaming over WebSocket and proxima commands in the terminal.

Good for

  • Asking several frontier models the same coding question from inside Cursor or VS Code
  • Running code review, test generation or security audits through MCP tools
  • Keeping source code on local hardware by routing requests to Ollama or LM Studio
  • Piping build errors or git diffs into an AI model from the terminal
  • Experimenting with a self-correcting local coding agent on a repository

Pricing

Freemium

Session mode is free without API keys; BYOK uses paid provider API keys; personal-use license only.

Prices change. Check proximamcp.in before you buy.

Proxima FAQ

Is Proxima free?

The README says session mode has no fees and needs no paid API keys, while BYOK mode uses the user's own API tokens, which the provider charges for.

Which platforms are supported?

The README lists Windows, macOS and Linux, with an installer for Windows and builds for the other systems available from GitHub releases or run from source.

What license does it use?

The README states a Proxima Personal Use License covering personal, non-commercial use only; commercial licensing requires contacting the author.

Does it send my code to the cloud?

The README says Proxima itself stores nothing externally, but session mode still sends prompts to the third-party provider being used, while local models keep code on the machine.

Which editors can connect to it?

The README names Cursor, VS Code, Windsurf, Antigravity and Claude Desktop, with the MCP config block shown for Cursor.

What are the best alternatives to Proxima?

The closest alternatives to Proxima are Local AI MCP, Opper, Unified AI System MCP Gateway, Agentic MCP and playwright_mcp.

Where this comes from

  • Description: written from the project's README and the project's website.

How AIPoweredBot checks servers