GitHub radar
Brigade: Open-Source AI Agent Crew With Persistent Memory
Spinabot released Brigade, an open-source framework for a personal AI agent that remembers facts across sessions and messages you back on Telegram, WhatsApp, and Slack. It runs on your own machine, with no cloud component or telemetry.
Brigade is an open source platform for building what its author calls “your personal intelligence at enterprise level”: a team of agents modeled after an actual org chart, with shared long term memory (what we call Tideline) and decaying facts / deduplication. Each agent can have different credentials and areas of focus, and there are rules around messaging so that they can delegate tasks to each other. There are 56 prebuilt skills, and it’s easy to integrate with Gmail, Notion, calendars, and thousands of other apps with just one key. Your agents will be able to read PDFs, Word docs, Excel files, and images that you send them, and they’ll respond with documents and reports of their own. Communicate with your agents via terminal, WhatsApp, Telegram, Slack, Discord, iMessage, BlueBubbles. Pick your favorite AI model: Claude, OpenAI, Gemini, OpenRouter, Groq, Mistral, DeepSeek, local Ollama, or any other OpenAI compatible server (even mix and match during conversations!). Built by Spinabot, under the MIT license. Try it out by asking your agent to deploy Brigade from github.com/spinabot/brigade. Follow the README instructions to install using your Claude Code credentials (or enter the API key it prompts you for), setup a Telegram bot, and verify that your agent responds correctly when you ask it a question and tell it a fact.
Why a vibe-coder should care
Rather than building your own bot quilt to have a single AI assistant for all your chat platforms that keeps track of conversation context, Brigade provides an out-of-the-box solution. Since everything is run locally, and there are no keys stored externally, I feel comfortable integrating my personal email/calendar with the single integration key. If you already have a subscription to Claude Code, you can use the same model without an additional fee. Or, if you want to save some money, you can use the free Ollama for basic repetitive tasks, although you probably shouldn’t expect high quality responses. Brigade works on a standard laptop without a GPU — the authors say the whole team runs even on a Raspberry Pi (you can even run your own local model through Ollama if you want, although you’ll need a GPU then).
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Deploy Brigade from this repository: https://github.com/spinabot/brigade — follow the README, use my Claude Code login during setup (or ask me for an API key for whichever provider), connect a Telegram bot, and show me that the agent answers and remembers a fact I tell it.
A regular laptop is more than enough — the authors say the whole crew runs even on a Raspberry Pi; no GPU needed unless you plug in your own local model through Ollama.
Open on GitHub▌ More finds