New · v0.1 early access

Your terminal's new AI partner.

Nexvexa Agent unifies OpenAI, Anthropic, Google and Ollama models behind a single terminal UI. Sandbox, budgets, RAG, macros and a Telegram bridge included.

$curl -fsSL https://nexvexa.pro/install.sh | bash

One terminal. Every backend.

OpenAIAnthropicGoogle GeminiOllamaCustom OpenAI-compatible

See it in 10 seconds

Try the live simulation before downloading: run terminal-style commands and walk through a dashboard preview.

Terminal Simulation
nexvexa-agent v0.1 · session:demo
$
Dashboard Walkthrough

Keys

OpenAI key: sk-**** · fp:7ca91e2a

Security: UNLOCKED

Automation

BTC below 60000 -> BUY (dry-run)

next run: 00:00:42

Trade Audit

audit verify OK · rows=28

manual BUY BTCUSDT qty=0.001 DRY ok

This is a marketing simulation and does not place real orders.

Everything in one app

Multi-model LLM control, security, automation and operational comfort.

Multi-backend

Hot-swap between OpenAI, Anthropic, Google, Ollama and OpenAI-compatible custom endpoints.

Sandbox & security

Path jail, project roots, profile-based approvals, audit log and one-click safe mode.

Telegram bridge

Run every slash command and chat from your Telegram bot.

Local RAG

Index a folder, query with embedding cache. Your data never leaves the box.

Macros

Bind frequent slash sequences to Ctrl+1-9 hotkeys.

Web dashboard

Local HTTP UI for stats, automations, sandbox events and CI webhooks.

Budget

Daily step and token caps; halt or prompt when exceeded.

Cold-start templates

Bootstrap projects in seconds with /template react-ts, python-api and more.

Undo

Every write/edit goes into a rollback queue; revert with /undo.

Layered security

Sandbox profiles, path jail, daily budgets and audit logs.

Path jail

Every file path the model writes is rejected unless it is inside the working directory or an explicitly approved root. Symlink escapes are blocked too.

Sandbox profiles

Predefined Dev / Trader / Content / SysAdmin profiles tighten network, file and binary access for each persona.

Audit log

Every tool call, approval and warning is written to a central log file. Bundle it with /export support-bundle.

Safe mode

One /safe on applies the strict profile, kills all network and forces approval on every action.

Ready to try it today?

Bring your own API keys or run fully local with Ollama.