Your Agent reads, writes, and thinks. Now it can hear. Flomem is a $69 open-source recording card that slips into your wallet and streams audio to your Agent as a native skill. No subscription. No vendor cloud. No lock-in.
Your Agent is blind to what you say.
It reads your text. It sees your code. It watches your screen.
But it can't hear your meetings, your voice notes,
your best ideas on a walk — the ones you lose
before you open your laptop.
BLE 5.2 handshake. Relay app runs silently in the background.
30s one-time setupWorks with Claude Code, any Agent with MCP / skill support.
/install flomemOne tactile press to start. Your Agent pulls the audio on demand.
/flomem fetch85.6 × 54 × 4 mm. The exact footprint of a credit card, thinner than two stacked coins. Slides into the card slot you already carry. No lanyard, no pendant, no orb stuck to your temple with medical tape.
Pay once. Use forever. No $10/mo transcription tier. No "premium" API quota. Your Agent already has API keys, models, and compute. Flomem just delivers the raw audio — that's it.
Firmware. Relay App. PC Agent. All on GitHub from day one. Fork it. Self-host the relay. Swap the STT provider. Your Agent's perception belongs to you.
/flomem fetch is a real skill your Agent calls like any other tool.
No webhook gymnastics. No "export → upload → prompt" loop.
Your Agent decides when to listen back.
Every "AI recorder" on the market charges you a subscription for transcription and AI — features your Agent already does better. Flomem strips the hardware down to what the silicon actually costs. Ears shouldn't cost more than the brain.
* Based on publicly-listed pricing at time of campaign launch. Omi's subscription is optional (self-host available, but setup is non-trivial).
We use Agents every day.
They read our code. They draft our emails. They ship our side projects.
But every time we had an idea in a meeting, on a call, on a walk — the Agent was deaf.
We didn't want another app with another subscription and another cloud we can't audit. We wanted our Agent to hear what we hear.
So we built it its own ears.
ATS2853 firmware in C. Toolchain, flash scripts, HAL for the mic array and BLE stack. Reflash with a USB-C cable and a stock toolchain.
iOS + Android background relay. React Native + native audio handlers. Swap the destination endpoint; point it at your own server if you want.
MCP server + reference skill files for Claude Code and OpenClaw. Includes a self-host guide so you can run the whole pipeline on a $5/mo VPS.
Pricing locked in for Kickstarter launch. Early Bird opens first — reserve your slot below to get the link the minute we go live.
/flomem fetch is a first-class skill your existing Agent (Claude Code, OpenClaw, whatever) calls natively. If Omi is a wearable assistant, Flomem is a USB mic for your Agent.
Schedule. Hardware slips. We've built buffer into Q4 shipping, but global logistics or a PCB revision could push a few weeks.
Agent ecosystem drift. MCP and skill formats are evolving. We'll maintain the skill for the top 3 agents for at least 2 years and keep the source open so the community can fork.
What isn't a risk. Hardware feasibility. The DVT works. We've manufactured similar-complexity audio devices at this exact factory. This is not a "can we build it" project — it's a "can we make enough" project.
Kickstarter campaign launching soon. Drop your email — we'll send you the Early Bird link the minute we go live.
●No spam. One email when we launch. Unsubscribe anytime.