Can M5Stack work Open Claw?

-- Views

April 27, 26

スライド概要

2026/04/27 M5 Japan Tour 2026 Spring Nagoya @ STATION Ai
https://m5stack.connpass.com/event/389541/

profile-image

Engineer / AWS Community Builder / LINE API Expert / JAWS UG名古屋 運営 GIthub: http://github.com/Miura55 Blog: http://supernove.hatenadiary.jp

シェア

またはPlayer版

埋め込む »CMSなどでJSが使えない場合

ダウンロード

関連スライド

各ページのテキスト
1.

Can M5Stack works Open Claw? Koki Miura(@k_miura_io)

2.

Introduction • Koki Miura • Engineer • Organizer • JAWS UG Nagoya & Kobe • Shachi.py • Orestudy @k_miura_io koki.miura05

3.

Notice • All slides in English for M5Stack team • This talk is in Japanese

4.

Maker Faire Shenzhen

5.

M5Stack Booth

7.

First Touch

9.

Can LLM8850 works Open Claw…?

10.

[Wed 2026-03-04 20:25 GMT+9] こんにちは Reply with ONLY the slug, nothing else. Examples: "vendor-pitch", "api-design", "bug-fix" [W][ Run][ 563]: history not append (rollback/modify). force ResetKVCache and recompute. [I][ SetKVCache][ 357]: prefill_grpid:2 kv_cache_num:128 precompute_len:0 input_num_token:6035 [I][ SetKVCache][ 359]: current prefill_max_token_num:1024 [I][ SetKVCache][ 360]: first run [I][ Run][ 412]: input token num : 6035, prefill_split_num : 48 malloc(): invalid size (unsorted) Aborted

11.

It Was Broken… • Raspberry Pi doesn’t recognize LLM8850 • LLM8850 doesn’t work... • Should I buy AI pyramid…?

14.

Yesterday https://x.com/M5Stack/status/2046945407050342835?s=20

15.

Wath’s esp-claw? • AI agent framework inspired by Open Claw • The implementation capabilities on ESP32, draws on MiniClaw https://github.com/espressif/esp-claw

16.

REQUIRE • ESP-Claw requires at least 8 MB Flash and 8 MB PSRAM • Supports ESP32-S3 and ESP32-P4

17.

Found!

18.

Let's gooo!!

19.

Burn Firmware

20.

Coneect & Setup

21.

Connected

22.

Architecture

23.

DEMO

24.

Movie

25.

Conclusion • M5Stack Core S3 can work generic Open Claw • Unless you slightly adjust the wording of your question, you won’t get the answer you’re looking for • Too many rate limit using Claude API • Try local AI agent using LLM8850 someday

26.

END