PROXIMA IS LIVE // V 5.5.0 READY

Proxima Multi-AI MCP Server:
UNIFY
Your WORKSPACE

> Run ChatGPT, Claude, Gemini & Perplexity inside Cursor, VS Code, Windsurf & Antigravity
Free Local Multi-AI MCP Server & Autonomous Action Agent — Zero API Key Costs

41+ MCP TOOLS
/
4 MODELS
/
1900+ USERS
COMMUNITY MILESTONES Independent Recognition

OPEN SOURCE RECOGNITION

Independent community rankings and trending milestones tracked on GitHub and Trendshift.

GitHub Trending Daily Overall
#3

#3 Repository Of The Day

Featured on GitHub Trending, first reaching #3 on May 7, 2026.

Trendshift Daily JavaScript
#13

#13 Repository Of The Day

Ranked on Trendshift's daily list for JavaScript, first achieved on 12th, Apr 2026.

Trendshift Weekly JavaScript
#15

#15 Repository Of The Week

Ranked on Trendshift's weekly list for JavaScript, first achieved in Week 16, 2026.

Three Powerful Ways to Run Proxima

Choose between 100% free browser sessions, your own custom API keys, or completely offline local hardware.

Session Routing

100% Free · Default

Automates your active browser logins inside sandboxed background views. Zero API keys, zero token fees, unlimited coding.

  • ChatGPT, Claude, Gemini & Perplexity
  • No API subscriptions or credit cards needed
  • Renderer-level WebSocket token stream interception
  • Headless background execution on localhost

Bring Your Own Key

BYOK · All Providers

Direct official API access for production pipelines. Encrypted securely in your operating system's native keychain vault.

  • OpenAI, Claude, Gemini, DeepSeek (V3/R1), Groq, xAI
  • OpenRouter, Together AI, Fireworks, Mistral, NVIDIA NIM
  • OS-native keychain encryption (Electron SafeStorage)
  • Auto Context Compaction & smart token pruner

Local Offline Models

100% Offline · Private

Run open-weight models on your local GPU or CPU. Your code never leaves your computer — zero internet required.

  • Ollama, LM Studio, vLLM & LocalAI compatible
  • Total data isolation and offline privacy
  • Works with Llama 3.3, Qwen 2.5, DeepSeek R1 local
  • Unified MCP tools and agent support for local models

Your Local AI Gateway

Route prompts to the best model for the task. Zero friction.

Stop Relying on a Single Model.

AI coding agents hallucinate, guess outdated APIs, and struggle with newer frameworks — because they're stuck with one model's old training data. Proxima lets multiple AI models with web access and stronger reasoning work together, cross-verify each other's answers, and produce results no single model can match alone.

$ proxima ask claude "review this code"
$ proxima fix "TypeError: undefined"
$ git diff | proxima code review

REST API & WebSocket

OpenAI-compatible endpoint at localhost:3210 (the port is auto-discovered if it is busy). Stream responses in real-time via WebSocket, set thinking depth, and cancel in-flight requests.

POST /v1/chat/completions
GET  /v1/models
POST /v1/cancel
WS   ws://localhost:3210/ws

41 MCP Tools

list_available_models, deep_search, review_code, security_audit, write_tests, fix_error, and 35 more — ready to use in Cursor, VS Code, Windsurf, Antigravity, or Claude Desktop.

Multi-AI Orchestration

ask_all_ais, smart_query, crew, run_workflow, debate — send one prompt across multiple AI providers at once, cross-verify answers, or chain outputs between role-based agents.

Terminal CLI

Run "proxima ask", "proxima fix", "proxima code" from any terminal. Supports pipes — send build errors or git diffs straight to AI.

Runs Local & Private

Everything on localhost. No cloud relay. Works headless in background — MCP tools keep running even without the app window open.

Autonomous Action Agent

Turns prompts into real-world action. Executes terminal tasks, creates scripts, manages files, and completes multi-step goals across your computer.

Automate Any Workflow

Automate virtually any task across your entire system — from browser actions and web research to custom scripts, API data pipelines, and background workflows.

Autonomous Proxima Agent

An autonomous local action engine that completes tasks, automates workflows, and handles complex computer operations hands-free — runs without API keys, with custom API keys, or completely offline with local models.

Hands-Free Task Completion

Takes your high-level goals and executes them end-to-end — running commands, managing files, processing data, and writing code.

Full Web & Browser Automation

Navigates the web, extracts live research, interacts with online services, and automates browser actions directly through Chrome.

Self-Correcting & Reliable

Automatically catches terminal errors, analyzes what went wrong, and fixes issues iteratively until the task is successfully finished.

Cross-Session Memory & Skills

Remembers your workspace preferences, learns from past workflows, and creates reusable automation shortcuts over time.

Proxima Agent Web UI localhost:8500
Proxima Agent Live Execution
SQLite Engines: vault.db insights.db memory.db skills.db
Modes: Auto · Smart · Suggest

What's New in Proxima 5.5.0

Pick models from what your accounts actually offer, control reasoning depth, cancel runaway requests, and stop worrying about busy ports.

From the Engineering Journal

Deep technical architectural guides, Model Context Protocol internals, and autonomous agent loops.

Watch Proxima MCP, REST API & CLI in Action

Experience seamless MCP tool calling, OpenAI-compatible REST endpoints, and terminal CLI workflows in real-time.

Full Demo
Live Chat
CLI Usage
App Overview

Model Agnostic Intelligence.

Switch between providers instantly. Use the best model for each task.

ChatGPT

Best for code generation, debugging, and structured output

  • Strong at coding tasks
  • Plugin & tool calling
  • Fast structured output

Claude

Best for deep reasoning, long code reviews, and analysis

  • Large context window
  • Careful, detailed reasoning
  • Best for refactoring

Gemini

Best for multimodal tasks, image understanding, and research

  • Image & video input
  • Massive context window
  • Google Search grounding

Perplexity

Best for real-time research, current docs, and citations

  • Live web search built-in
  • Source citations on every answer
  • Latest framework docs
EXTENDED BYOK & LOCAL HARDWARE

Also Supports Your Custom API Keys & 100% Offline Models

DeepSeek (V3/R1) Groq (Ultra-Fast) xAI (Grok 2) Ollama (Offline) LM Studio Mistral OpenRouter

41 MCP Tools at Your Fingertips

Everything you need for AI-powered development. The list below is the exact set the MCP server registers in v5.5.0, plus the proxima://status and proxima://models resources.

list_available_models Discover live models per account
ask_chatgpt Query ChatGPT
ask_claude Query Claude
ask_gemini Query Gemini, with thinking depth
ask_perplexity Query Perplexity (web + citations)
ask_model Any session or BYOK provider
ask_all_ais Same prompt to every provider
smart_query Auto-route: verify, consensus, collaborate
new_conversation Reset a provider's context
generate_code Write new code from a description
explain_code Line-by-line explanations
review_code Bugs, security, performance review
optimize_code Speed, memory, readability
fix_error Root cause and exact fix
explain_error Plain-language error help
verify_code Best-practices check
convert_code Translate between languages
write_tests Generate a full test file
security_audit Severity-ranked vulnerability audit
build_architecture Design a system architecture
solve End-to-end coding task solver
content Summarize, write, brainstorm, analyze, extract
compare Side-by-side comparison
debate Providers argue different stances
verify Cross-check a claim with a confidence rating
crew Role-based multi-model pipeline
run_workflow Ordered multi-step pipeline
run_loop Generate, review, improve until converged
proxima_cost_report Token usage and estimated savings
proxima_agentic_status Router, memory and health diagnostics
analyze_file File or whole-folder analysis, secret scan
review_code_file Review a single file on disk
clear_cache Drop cached provider responses
show_window Show the Proxima window
hide_window Hide the window, MCP keeps running
toggle_window Toggle window visibility
set_headless_mode Run in the background

Works With Your Favorite Tools

Cursor
VS Code
Claude Desktop
Windsurf
Antigravity
Any MCP Client

Up and Running in Minutes

Simple setup, powerful results

1

Download & Install

Download the installer from GitHub releases (Windows installer; macOS and Linux builds are available too), or run from source with npm install and npm start.

Download Latest Release →
2

Login to Providers

Open Proxima and sign in to your AI accounts (ChatGPT, Claude, etc.)

3

Copy MCP Config

Go to Settings and copy the MCP configuration JSON.

mcp-config.json
{
  "mcpServers": {
    "proxima": {
      "command": "node",
      "args": ["path/to/proxima/src/mcp/index.js"]
    }
  }
}
4

Start Using

Paste config into your AI coding app (Cursor, VS Code, etc.) and enjoy!

Stop Paying for Multiple Subs.

Proxima unifies your existing accounts. Switch models instantly. Build faster.

Proxima Agent Full View
Proxima Agent — Live Execution View