Amurex

Open-source AI meeting assistant, honestly reviewed. Self-hostable, privacy-focused, and genuinely rough around the edges.

TL;DR

• What it is: Open-source (AGPL-3.0) AI meeting copilot delivered as a Chrome extension. Transcribes meetings, generates summaries, surfaces real-time suggestions, and connects to your knowledge tools — Notion, Google Drive, Obsidian [1][4].
• Who it’s for: Solo founders and small teams who live in Google Meet or MS Teams and want meeting notes automated without paying for Fireflies, Otter.ai, or Granola subscriptions [2][4].
• Cost savings: Otter.ai Pro runs $16.99/mo. Fireflies Business is $19/mo per seat. Amurex self-hosted costs you a VPS and an afternoon — assuming you can navigate the setup [1][4].
• Key strength: Genuinely open-source (AGPL-3.0), fully self-hostable, works as an invisible browser layer without installing a separate desktop app. Privacy-first by design [1][2][4].
• Key weakness: Only works on Google Meet and MS Teams — no Zoom. Setup requires Supabase, Redis, and Python 3.11, which is a meaningful technical bar. The project has 2,823 GitHub stars but monthly web traffic dropped 54.5% to 4.1k visits — a concerning signal for a tool betting on community momentum [1][2].

---

What is Amurex

Amurex is a Chrome extension paired with a self-hostable Python backend that acts as what the project calls “an invisible companion” during your meetings [1]. You install the extension, connect it to your knowledge sources (Notion, Google Drive, Obsidian), and it starts transcribing, summarizing, and suggesting — without you switching tabs or launching a separate app.

The project is built by The Personal AI Company and sits on GitHub under thepersonalaicompany/amurex-backend, licensed AGPL-3.0. The main front-end repository has accumulated 2,823 stars and 219 forks as of this review [2]. The backend repo is smaller and less documented, but it’s the piece you need if you want to actually self-host.

What Amurex is trying to do is stitch together two separate categories that usually don’t talk to each other: meeting intelligence (transcription, summaries, follow-ups) and knowledge management (unified search across your connected tools). Most tools pick one. Amurex claims both [1][4].

The Chrome extension angle is smart — it means there’s nothing to install on the desktop, no Electron app, no system permissions for audio capture beyond what the browser already has. If you’ve used Granola or Fireflies, the meeting capture experience is conceptually similar, but the delivery mechanism is different and the privacy story is stronger when you’re running your own backend [1][2].

---

Why People Choose It

The reviews we have access to are thin — there are no Trustpilot or G2 rating pools for Amurex, and AlternativeTo has zero user comments or reviews as of this writing [2]. What we can synthesize comes from tool directory writeups and the product’s own feature descriptions.

Versus Otter.ai, Fireflies, and Granola. These are the direct comparisons. Otter, Fireflies, and Granola are all proprietary SaaS with monthly seats. Your transcripts live on their servers. If you’re running a founder meeting where IP or strategy is on the table, that’s a real concern. Amurex’s pitch — full self-hosting, local processing optional — directly addresses this [1][4]. The tradeoff is that those tools have polished, stable products with multi-platform support (including Zoom). Amurex doesn’t.

The privacy angle. BestAIToolFinder’s writeup [4] leans hard on data sovereignty: “Maintain full data sovereignty by hosting Amurex on your own infrastructure, ensuring complete control over your proprietary data.” The backend README confirms this is real, not just marketing — you can run in CLIENT_MODE=LOCAL which swaps OpenAI/Groq for local Ollama inference and local embeddings via fast-embed. That means the entire pipeline — transcription, summarization, embedding — can run air-gapped on your own hardware [README].

The knowledge integration angle. Most meeting tools stop at transcription. Amurex connects to Notion, Google Drive, and Obsidian and builds a searchable knowledge base across all of them [1][4]. In theory, you can ask “what did we decide about pricing in last month’s calls?” and it knows — because it has both your transcripts and your docs indexed. In practice, whether this works smoothly depends heavily on setup quality, and documentation here is thin [1].

What’s less convincing. The -54.5% traffic drop to 4.1k monthly visits [1] is hard to explain away. Either the broader AI tools market is fragmenting fast (true) or something about Amurex specifically is losing users (also possibly true). A tool with declining traffic and no published user reviews is a signal worth noting before you build your meeting workflow around it.

---

Features

Based on the README, aipure.ai [1], and AlternativeTo [2]:

Meeting intelligence:

• Real-time AI suggestions during calls, context-aware to the current conversation [1][2]
• Automatic transcription with smart summaries and key takeaways [1][2]
• Late-join recap — if you join a meeting 15 minutes late, you get a catch-up summary without interrupting the meeting [1][2]
• One-click follow-up email generation after the meeting ends [1][2]
• Works on Google Meet and MS Teams; no Zoom support currently [1]

Knowledge management:

• Unified search across Notion, Google Drive, Obsidian [1][4]
• Context-aware natural language queries across your connected tools [4]
• Cross-tool knowledge aggregation — builds a knowledge graph from your indexed content [4]

Privacy and infrastructure options:

• CLIENT_MODE=ONLINE: uses OpenAI, Groq, Mistral APIs for inference and embeddings [README]
• CLIENT_MODE=LOCAL: uses local Ollama for inference and fast-embed for embeddings — fully air-gapped [README]
• Full self-hosting via Docker or Docker Compose [README]
• AGPL-3.0 license — source code fully available [1][2]

What’s notably absent:

• No native Zoom or Webex support — a significant gap for teams not on Google Meet/MS Teams [1]
• No mention of mobile apps
• No calendar integration or automated meeting join
• No mention of speaker diarization (identifying who said what)

---

Pricing: SaaS vs Self-Hosted Math

Amurex doesn’t publish a public pricing page — data not available from scraped sources (the website returned a 503 during research). Based on available evidence, the cloud version appears to be free to use, and the self-hosted version is free software [1][2].

What comparable tools charge:

• Otter.ai Pro: $16.99/mo (300 monthly transcription minutes, AI features)
• Fireflies Business: $19/mo per seat
• Granola: $18/mo
• Fathom: free tier available, paid plans from $15–19/mo per user

Self-hosted Amurex:

• Software: $0 (AGPL-3.0)
• VPS to run it: $6–10/mo (the backend is Python, needs Redis + Supabase-compatible Postgres)
• External API costs if running in ONLINE mode: OpenAI/Groq charges per token — variable, but light for meeting transcription workloads
• LOCAL mode: $0 additional, but requires Ollama running somewhere accessible

For a two-person founding team paying Fireflies at $38/mo combined, self-hosting Amurex on a $6 VPS and connecting it to a free Groq API key would realistically cost $6–10/mo. That’s $28–32/mo saved — roughly $340–380/year.

The catch is setup time and maintenance. If this breaks in three months when the project updates its Supabase schema, you’re debugging Python dependencies on a Saturday afternoon.

---

Deployment Reality Check

This is where Amurex gets honest about who it’s really for. The README setup path requires [README]:

• Python 3.11 (exact version — 3.12 may not work)
• A Supabase project (you create the project, run migration SQL, configure storage buckets)
• Redis (username, password, host, port — not just a local default instance)
• Resend account for email (API key + noreply address)
• At least one of: OpenAI key, Groq key, Mistral key (or Ollama running locally)
• Docker for the container path

That’s five external dependencies before you’ve written a single line of code. The Supabase setup alone requires creating tables via SQL migration scripts and configuring CORS on storage buckets [README].

aipure.ai [1] calls this out directly: “Complex setup process for offline/self-hosted deployment” and “potentially steep learning curve due to limited documentation.”

If you’ve deployed a Docker Compose stack before and you’re comfortable with .env files, this is probably two to four hours. If you’ve never touched a Linux server, this is a weekend project with a realistic chance of failure due to gaps in the documentation.

One quirk worth flagging: the README includes a manual caution in all-caps that if running locally you need to comment out fastembed==0.4.2 from requirements.txt before installing [README]. That kind of manual workaround warning in the primary install path is a signal about documentation maturity.

The LOCAL mode (Ollama + fast-embed) adds another layer: you need Ollama running and accessible at a known endpoint. This is excellent for privacy but doubles the infrastructure surface area.

Realistic time estimates:

• Technical founder comfortable with Docker/Supabase: 2–4 hours
• Non-technical founder following a guide (if a good one exists): probably not feasible without help

---

Pros and Cons

Pros

• Genuinely open-source. AGPL-3.0 means you can inspect, modify, and self-host the entire stack. No “open core” with a proprietary cloud layer [1][2].
• Full self-hosting with local AI. The CLIENT_MODE=LOCAL path with Ollama means zero data ever leaves your infrastructure — a real privacy story, not a marketing one [README][1].
• Invisible browser layer. Chrome extension model means no desktop app to install, no system-level audio hooks, nothing that requires elevated permissions beyond the browser [1][2].
• Unified knowledge search. The combination of meeting transcripts + connected knowledge tools (Notion, Drive, Obsidian) in one search surface is genuinely useful for founders drowning in scattered context [1][4].
• No per-seat or per-minute pricing (from all available evidence) — free software running on your own VPS [1][2].

Cons

• Google Meet and MS Teams only. No Zoom. This immediately disqualifies it for a large portion of the market [1].
• Complex setup for a solo founder. Requires Supabase project creation, Redis, Python 3.11, and at least one AI API key. Not a docker run one-liner [README][1].
• Thin documentation and community. No visible user reviews on Trustpilot or G2. Zero comments on AlternativeTo despite being listed since December 2024 [2]. Limited docs flagged by multiple sources [1].
• Traffic declining sharply. A -54.5% drop to 4.1k monthly visits [1] suggests this tool isn’t growing — which matters for open-source tools that depend on community contributions and issue resolution.
• AGPL-3.0, not MIT. If you want to embed Amurex into a commercial product or SaaS offering, AGPL requires you to open-source your modifications and the integrated system. This is a real legal constraint that n8n (Fair-code) and MIT tools don’t have the same way [2].
• No speaker diarization mentioned. The transcription feature doesn’t appear to identify who said what — a common gap in simpler meeting tools.
• Manual install quirks (like the fastembed comment-out requirement) signal this codebase isn’t yet polished enough to install without reading the README carefully [README].

---

Who Should Use This / Who Shouldn’t

Use Amurex if:

• You’re primarily on Google Meet or MS Teams — both personally and with clients.
• You’re paying $15–20/mo for Otter or Fireflies and the privacy concern is real (strategy calls, legal discussions, client NDA territory).
• You have basic Docker experience or someone on your team who does.
• You want to connect meeting intelligence to your existing Notion/Drive/Obsidian knowledge base rather than building a separate silo.
• You’re willing to accept some setup friction and rougher edges in exchange for data sovereignty and $0 SaaS bill.

Skip it if:

• You’re regularly on Zoom calls — there’s no support [1].
• You’re not technical and don’t have a technical partner or contractor who can deploy and maintain this.
• You need a tool that’s stable, documented, and has a real support path — the declining traffic and absent community reviews are warning signs for anyone making a serious operational bet.
• You want to embed or resell this capability in your own product — the AGPL-3.0 license creates obligations that MIT or Apache-2.0 tools don’t [2].

Stay on Otter or Fireflies if:

• Your meetings are on multiple platforms (Zoom + Google Meet + Teams).
• You need speaker identification and multi-speaker transcripts.
• Reliability matters more than privacy and cost to you.

---

Alternatives Worth Considering

• Otter.ai — most mature AI transcription tool, multi-platform, proprietary SaaS. Starts at $16.99/mo. Best if you want something that just works across all meeting platforms.
• Fireflies.ai — stronger on CRM integrations and team collaboration on transcripts, proprietary, $19/mo per seat. Good for sales teams.
• Granola — desktop-only, Mac-first, elegant UI, proprietary, $18/mo. Very good for note quality; no self-hosting.
• Fathom — free tier is genuinely usable, Zoom-first, proprietary. Best free option if Zoom is your primary platform.
• Notionite / Rewatch / Loom — depend on your use case (documentation vs. async video vs. meeting capture).
• Open-source alternatives: There isn’t a well-maintained, widely-deployed open-source meeting assistant with comparable features. Amurex is essentially alone in this space, which is both its opportunity and its risk.

---

Bottom Line

Amurex makes a credible pitch for one specific profile: a privacy-conscious founder who lives on Google Meet or MS Teams, has basic Docker skills, and wants to stop paying $15–20/mo for meeting transcription SaaS. The local AI mode is genuinely private, the knowledge integration angle is thoughtful, and the Chrome extension delivery is smarter than requiring a desktop app.

But the warning signs are real. A -54.5% traffic drop, no visible user reviews anywhere, complex multi-dependency setup, and documentation gaps all point to a project that’s earlier and rougher than its feature list implies. If you’re betting your meeting workflow on a tool, you want one with community momentum, not one with declining traffic and manual workarounds in the install README.

If the setup friction is the blocker, that’s what upready.dev handles for clients — one-time deployment, you own the infrastructure, no ongoing SaaS bill.

---

Sources

1. AIPure.ai — Amurex: Reviews, Features, Pricing, Guides, and Alternatives (aipure.ai). https://aipure.ai/products/amurex
2. AlternativeTo — Amurex: Transform your meeting experience with AI intelligence (alternativeto.net). https://alternativeto.net/software/amurex/about/
3. BestAIToolFinder — Amurex (bestaitoolfinder.com). https://bestaitoolfinder.com/amurex/

Primary sources:

• GitHub backend repository: https://github.com/thepersonalaicompany/amurex-backend (AGPL-3.0)
• Official website: https://www.amurex.ai
• Chrome Web Store listing: https://chrome.google.com/webstore/detail/amurex/dckidmhhpnfhachdpobgfbjnhfnmddmc