Jarvis (Personal Jarvis) is a free, open-source desktop AI assistant you operate by talking to it. You say the wake word, speak a request, and it answers or acts: running your agents, coding CLIs, browser, and apps across macOS, Windows, and Linux. Voice is handled locally first, and the language model behind it is whichever provider you bring your own key for. It serves individuals who want hands-free control of an AI setup they already use. It is a voice-first, open-source AI assistant for the desktop, distributed from personaljarvis.ai with its source at PersonalJarvis/PersonalJarvis under the Apache-2.0 license. The project is young at 168 stars and 12 forks observed in October 2026, with a second generation on a waitlist; the constant is the shape of the promise: your voice, your model keys, your machine.[1][2][3]
Quick verdict
People who already direct Claude Code or Codex from a terminal benefit most from Jarvis, because it puts voice operation on top of those same agents. Its 40-plus slash commands, custom commands, and spaced-repetition learning keep repeated requests short, while MCP protocol support connects it to outside tools. The catch is maturity: with 168 stars observed in October 2026 and a second generation on a waitlist, this is an early-stage project, not a polished consumer product. If you want hands-free agent operation and accept early software, try Jarvis; if you want a settled desktop assistant today, look at 5ire instead.[1][2][3]
Features that matter
Wake-word voice control Wake-word voice control: saying Jarvis starts listening, and spoken commands run locally first.[1]
Local-first speech pipeline Local-first speech pipeline: voice recognition through faster-whisper with replies through Kokoro, Piper, or Chatterbox voices.[1]
Agent orchestration Agent orchestration: Jarvis runs Claude Code, Codex, and mixed teams of agents from one spoken request.[1][2]
Command of your software Command of your software: operation of coding CLIs, the browser, and desktop apps from speech.[1][2]
Slash and custom commands Slash and custom commands: more than 40 built-in slash commands plus custom commands for repeated personal requests.[1][2]
Spaced-repetition learning Spaced-repetition learning: scheduled revisits so repeated preferences and facts persist.[1]
MCP protocol and automations MCP protocol and automations: protocol support plus workflow automations connecting outside tools and routines.[1][2]
Your models, your keys Your models, your keys: bring-your-own-keys access to OpenAI, Claude, Gemini, and open-source models, in English or French.[1][2]
Where Jarvis fits
Hands-free control of a desktop computer and its AI agents: the wake word plus local-first voice handling keeps spoken requests on the machine.[1][2]
Developers already using Claude Code or Codex who want to operate them by voice instead of typing.[1][2]
Privacy-minded individuals who prefer local-first voice handling with model keys they control.[1][2]
Where it does not fit
Buyers needing a managed cloud service with SSO, audit, or a support plan: none of these is documented.[1][2]
Buyers wanting polished non-technical onboarding from a mature product with a large user base: 168 stars observed signals early software.[1]
Mobile-only use: distribution is desktop installers for macOS, Windows, and Linux, with no mobile app evidenced.[1]
Alternatives and compatibility
Choose Jarvis when the deciding requirement is talking to your setup: voice-first operation of agents, CLIs, browser, and apps, free and open-source, with data on your machine, tolerating early software.[1][2][3]
If hands-free control is optional and you would rather have a settled assistant now, 5ire is the safer catalogue alternative, with a published record as a desktop AI assistant carrying MCP tools and a local knowledge base.[1][2]
Questions worth asking
Is Jarvis free to use?
Does Jarvis work offline?
Partly, and the boundary matters. Voice recognition and speech run through a local-first pipeline on your machine. The language model itself comes from your keys, so fully offline use depends on pointing Jarvis at a local model; the official pages do not promise a bundled offline LLM, and this assessment does not claim one.[1]
What is the best Jarvis alternative for a settled desktop assistant?
5ire, the incumbent in this catalogue relationship. It is a published open-source desktop AI assistant that chats with many providers and connects file, database, and remote-data tools through MCP servers, plus a local knowledge base. Pick 5ire for a mature chat-centered assistant; pick Jarvis when voice-first operation of agents is the requirement.[1][2]
Which platforms does Jarvis support?
macOS 12 and later, Windows 10 and 11, and Linux across Ubuntu, Debian, Fedora, and Arch, via per-OS desktop installers. English and French are the documented languages. No mobile app is evidenced.[1]
Sources and editorial review
Source research. Reviewed on October 11, 2026. Written by Sanjib Sah. AI assisted; reviewed by a human editor.
We have not run or tested this product.
- Personal Jarvis homepage (official)
official · Observed October 11, 2026
Official homepage: voice-first positioning, supported platforms, agent/browser/app operation, slash commands, MCP, BYOK, local data, free and open-source, gen-2 waitlist.
- Personal Jarvis documentation (official)
official · Observed October 11, 2026
Official docs: product definition, per-OS installers and versions, 3-step voice setup, local speech pipeline components, orchestration targets, commands, learning, MCP, automations, BYOK, FAQ.
- PersonalJarvis/PersonalJarvis repository
official · Observed October 11, 2026
Repository identity: name, Apache-2.0 license, star/fork counts, canonical links to site, docs, and Discord.
Compared with these products
No explicit proprietary baseline is recorded.