Flue Harness Under the hood - https://t.co/4vuSjfMaRB
As repo is changing fast - am keeping flue agent to refreshing this blog every few days.
New Deep Dive is coming soon....
Thanks @FredKSchott for accepting some of my work while learning this and there is lot on the way!
if you’re in India, subscribing to @opencode just got a lot easier
upi autopay is now live for opencode go
₹900/month for generous limits and reliable access to open source coding models
"Generated" desktop app to read all my markdown documents from one App - No need to open vscode/intelij to read "docs" directory : https://t.co/T0hgzcTOH7
Useful as I spend more time reading that coding
Supports Mermaid diagrams
It will build for MacOs and windows as well.
this week small wins
1. tacit desktop now streams claude sessions from remote machines (beta)
2. tacit now reads codex, gemini and claude sessions and groups them for project/repo/dir
3. small PR to new agent harness
Introducing Flue — The First Agent Harness Framework
Flue is a TypeScript framework for building the next generation of agents, designed around a built-in agent harness.
Flue is like Claude Code, but 100% headless and programmable. There's no baked in assumption like requiring a human operator to function. No TUI. No GUI. Just TypeScript.
But using Flue feels like using Claude Code. The agents you build act autonomously to solve problems and complete tasks. They require very little code to run. Most of the "logic" lives in Markdown: skills and context and AGENTS.md.
Flue is like Astro or Next.js for agents (not surprising, given my background 🙃). It's not another AI SDK. It's a proper runtime-agnostic framework. Write once, build, and deploy your agents anywhere (Node.js, Cloudflare, GitHub Actions, GitLab CI/CD, etc).
We originally built Flue to power AI workflows inside of the Astro GitHub repo. But then @_bgiori got his hands on it, and we realized that every agent needs a framework like Flue, not just us.
Check it out! It's early, but I'm curious to hear what people think. Are agents ready for their library -> framework moment?
Seeing is believing.
Just exported an old Claude Code session, - 𝗳𝘂𝗹𝗹 𝘁𝗵𝗿𝗲𝗮𝗱, 𝗲𝘃𝗲𝗿𝘆 𝘀𝘂𝗯𝗮𝗴𝗲𝗻𝘁 𝗼𝘂𝘁𝗰𝗼𝗺𝗲, 𝘁𝗵𝗲 𝘄𝗵𝗼𝗹𝗲 𝗿𝗲𝗮𝘀𝗼𝗻𝗶𝗻𝗴 𝘁𝗿𝗮𝗶𝗹 — 𝗶𝗻𝘁𝗼 𝗮 𝘀𝗶𝗻𝗴𝗹𝗲 𝘀𝗵𝗮𝗿𝗲𝗮𝗯𝗹𝗲 𝗳𝗶𝗹𝗲.
Why it matters:
→ Hand it to a teammate. They open it. They're caught up. No standup needed.
→ Embedded agent-friendly version inside, so another machine or another Claude can pick up exactly where I left off.
→ I literally sent one to a colleague over WhatsApp to explain what we'd built. He read it on his phone.
The shift: AI sessions stop being ephemeral chats and start being durable artifacts. Reviewable. Shareable. Resumable.
Demo (open in any browser): https://t.co/hKoUGNooVr
This is what session memory should feel like.
Building "easy" software got cheaper, "good enough" software faster.
Building "product" is still as easy or complex it used to be depending on hat you put on.
Bimodal mode is freaking and wearing out us builders faster for sure :)
@bentlegen I made claude code skill for excalidraw - https://t.co/vLbg8zqisT
A Claude Code skill that opens an Excalidraw drawing canvas, lets you sketch a diagram, and automatically attaches the exported PNG as visual context for your prompt.
Tested only for linux.