Back to directory

Jarvis AI Desktop Agent

A self-hosted, multi-LLM autonomous agent for Linux that controls a real desktop over VNC, receives tasks through web chat or WhatsApp, and automates browsing, file, email, spreadsheet and calendar work behind a sandboxed security layer.

Jarvis AI Desktop Agent

Jarvis is a self-hosted autonomous AI agent that runs on a Linux server and carries out a goal given in plain language through a web chat, a built-in support portal, a task pane inside Outlook or Excel, or WhatsApp. It plans and executes multi-step work: browsing the web, reading and writing files, running code, editing spreadsheets, generating Office documents and diagrams, answering email by rule, evaluating SAP data read-only, and managing a calendar, with an optional live VNC view so a user can watch or take over the desktop at any time.

It supports multiple LLM providers, including Google Gemini and Anthropic Claude models, switchable without restarting, and is built for multi-user use with a security layer that includes sandboxed execution, attack prevention and an incident log visible in its settings. Other documented capabilities include a multi-agent system, a skill system, a knowledge base, vision and face recognition, and AD/LDAP-based access control.

It is written in Python with FastAPI, and licensed under Apache-2.0.

More in Agent frameworks

See all 902
  • Recurse

    recurse.run

    A serverless harness that lets a coding agent design, test, and deploy its own specialist sub-agents as tools, MCP servers, or bots, defined by a schema-validated manifest.

  • ProofForge

    shulgin.is-a.dev

    An AI-agent pipeline that decomposes a mathematics problem, proves each piece, and formalizes the result in Lean 4 and Mathlib until the Lean kernel accepts it, with six of its proofs already merged into Google DeepMind's formal-conjectures repository.

  • local-coder

    github.com

    Builds a local coding-agent team on top of Ollama and OpenCode by detecting the machine's hardware, recommending models for seven specialist roles, and testing whether each assigned model can actually use tools, edit a repository and pass verification.