jarvis

Jarvis is a locally-run, voice-first AI assistant for a user’s computer. It listens for a wake word (“Jarvis”), performs speech recognition (Whisper), runs an on-device intent judge and a selectable LLM via Ollama, supports local tools (e.g., screenshot/OCR, web search, weather, file access, nutrition/location features), and can connect to external capabilities via MCP servers (e.g., GitHub, Home Assistant, Google Workspace, Slack/Discord, databases). It claims 100% local processing and automatic redaction, with conversation/memory stored locally.

Evaluated Mar 30, 2026 (21d ago)
Repo ↗ Ai Ml ai assistant voice local-first privacy mcp ollama productivity automation health machine-learning
⚙ Agent Friendliness
42
/ 100
Can an agent use this?
🔒 Security
30
/ 100
Is it safe for agents?
⚡ Reliability
24
/ 100
Does it work consistently?

Score Breakdown

⚙ Agent Friendliness

MCP Quality
55
Documentation
65
Error Messages
0
Auth Simplicity
55
Rate Limits
0

🔒 Security

TLS Enforcement
20
Auth Strength
35
Scope Granularity
20
Dep. Hygiene
40
Secret Handling
35

Strengths: README claims 100% local/offline processing, automatic redaction of sensitive info, and local storage under ~/.local/share/jarvis; also provides a privacy hardening config example disabling web search and location. Risks/unknowns: No documented central security model for Jarvis, token storage strategy, encryption at rest, or logging redaction guarantees. MCP token handling is via environment variables in examples; this can be safe if carefully managed but risks leaking secrets through logs/process listings depending on implementation. Web search and MCP integrations can introduce external data flows depending on configuration (not fully specified). TLS is not applicable to Jarvis itself as it is local, but MCP/proxy examples may involve network endpoints; TLS requirements are not described.

⚡ Reliability

Uptime/SLA
0
Version Stability
45
Breaking Changes
20
Error Recovery
30
AF Security Reliability

Best When

You want an offline-capable, local voice assistant that can also call local and third-party MCP tools on your own machine for personal automation and assistance.

Avoid When

You need a standardized remote API/SDK with guaranteed uptime, or you cannot run/maintain local model tooling and keep tokens/secrets secure on the host.

Use Cases

  • • Hands-free voice assistant for coding help and debugging while viewing on-screen errors (via screenshot/OCR and MCP tools)
  • • Contextual, ongoing spoken conversations with a short rolling context and long-term local memory
  • • Local personal productivity and planning (business/life coach tone)
  • • Health/nutrition goal tracking and meal-related assistance (claims local storage)
  • • Smart home control through MCP integration (e.g., Home Assistant)
  • • Developer/ops assistance with GitHub issues/PRs/workflows via MCP
  • • Retrieval/search and summarization via web search and/or MCP-connected knowledge sources

Not For

  • • Browser-based/text-chat-only use (README states voice-only for now)
  • • High-assurance/compliance-sensitive workflows without reviewing code (health advice, data handling, and “redaction” are not specified in technical detail)
  • • Environments requiring a hosted API/service endpoint with enterprise SLA guarantees
  • • Users who cannot run local models and dependencies (Ollama, voice components, GPU/RAM requirements)

Interface

REST API
No
GraphQL
No
gRPC
No
MCP Server
Yes
SDK
No
Webhooks
No

Authentication

Methods: Per-MCP-provider tokens via environment variables (e.g., GITHUB_TOKEN, API_ACCESS_TOKEN, GOOGLE_CLIENT_SECRET, SLACK tokens, COMPOSIO_API_KEY). Local-only operation; no documented auth for Jarvis itself in the README.
OAuth: No Scopes: No

Authentication is primarily delegated to individual MCP servers/connectors via env vars. The README provides examples but does not document a centralized auth model, token storage strategy, or scope granularity for Jarvis.

Pricing

Free tier: No
Requires CC: No

License section indicates personal use is free forever; commercial use requires contacting the author. No subscription pricing for the software itself is described.

Agent Metadata

Pagination
none
Idempotent
False
Retry Guidance
Not documented

Known Gotchas

  • ⚠ Voice-only interface: agents that expect a text/chat API may not have a direct integration path (README says no text chat UI yet).
  • ⚠ MCP integrations rely on external MCP server implementations and tokens; failures may look like tool errors rather than Jarvis-specific errors.
  • ⚠ Stop/interrupt behavior may be imperfect (README: “Stop” commands during speech sometimes get filtered as echo).
  • ⚠ Under-active development; primary development on macOS, Windows/Linux support may lag.

Alternatives

Other local voice assistants (e.g., Whisper-based pipelines + local LLMs) Home Assistant voice assistants with cloud/onsite integration Open-source agent frameworks that support MCP with separate UI/voice components Hosted voice assistants if offline guarantees are not required

Full Evaluation Report

Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for jarvis.

AI-powered analysis · PDF + markdown · Delivered within 30 minutes

$99

Package Brief

Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.

Delivered within 10 minutes

$3

Score Monitoring

Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.

Continuous monitoring

$3/mo
API endpoint ↗ Agent guide ↗ Report inaccuracy

Scores are editorial opinions as of 2026-03-30.

8642
Packages Evaluated
17761
Need Evaluation
586
Need Re-evaluation
♥
Community Powered