Muvon

Your best AI prompts aren't typed. Hold a key → speak → screenshot auto-pastes with your voice into Octomind, Claude Code, Cursor, or Codex. Meetings transcribed with summaries. Notes captured and cleaned up. All local. No subscription. Built for Mac getvext.app

@A_excursionist 👀

@muvonteam something great, TBA

Builders never sleep. What are you building? 👇

@buildfast_dev Hey! Kind of faster bootstrapping? You use AI there? Like boilerplate and things are great start for clean code. But later it's great to have some clear instruction to act on given architecture with AI

@muvonteam wassup! 👋 Building dev tools for faster shipping: - SaaS boilerplate — Next.js, auth, DB wired up - Stripe guide — add payments in 30 min Full-stack · freelancing · indie hacking nathanite3.gumroad.com

@mytwillot That's cool! Can search with semantic query also?

@muvonteam Still at it! Building a way to rescue your bookmarks and likes from the void—Twillot backs them up with AI classification and local search so they become a searchable knowledge base. twillot.com/en

@FUCORY What is it?

@muvonteam Use incur so you can just support both

Stop building MCP servers. Build CLIs that agents can use

@Type_Whisper Sounds neat! Thanks for explanation. We will take a look and give a try🙏🏼

@muvonteam Fair push. No formal benchmark post from me yet. But TypeWhisper is not cloud-only and not monthly-sub-only: you can run it fully local if you want, and it is GPL/open source with commercial licensing for proprietary use.

Vext 1.1 is out — voice-to-text for Mac that actually understands AI workflows. Hold a key. Speak. Your cleaned prompt appears at the cursor. Screenshot + voice together. Local only. No cloud, no sub, no account. $24.50 once. 50% off with VEXT50. → muvon.io/blog/introduci…

Everyone's chasing bigger models. We tested it: same client, same prompt, different retrieval. 60% token reduction. No model change. The real bottleneck? Context quality, not model size.

@Type_Whisper Did you run benchmarks to compare? we also supports long form writing. No cloud. Totally yours. Privacy first. Blazing fast. Data only on your disk. You do the same? Or its simple cloud on monthly basis payment?

@muvonteam tbh, a lot of dictation tools feel similar until longer-form writing is involved. TypeWhisper is stronger on engine choice, cleanup, and workflow tuning: typewhisper.com

4/ We built this because we needed it. Free trial: getvext.app or brew install muvon/tap/vext Full details: muvon.io/blog/introduci…

3/ Vext closes the gap. Hold key → speak → release. Clean text at your cursor. Screenshot + voice in one shot. Local processing only. No cloud. No sub. $24.50 once.

Build tools for humans first. If the interface is clean, agents figure out the rest. Octocode — structural code intelligence. CLI first. MCP included. Rust. Open source. One binary.

Then I gave Claude access to it. claude mcp add octocode -- octocode mcp Claude could suddenly navigate my codebase — follow imports, trace dependencies, understand structure.

Octocode: brew install muvon/tap/octocode octocode index octocode search "authentication middleware" It parses your codebase with tree-sitter AST. Builds a knowledge graph. You search it. That's the tool.

The difference shows up in the first 30 seconds. Agent-first tools: install Python runtime, configure an index, spin up a server, wire it to your agent through a JSON file, hope the schema matches.

Most MCP servers for code search were built for agents. Octocode was built for us.