Orca Bots Now Check Your Projects Every 4 Hours on Their Own, With 24 Templates to Start From
Orca, a development environment for AI agents, now has a Bots tab. A bot doesn't wait for your prompt: every 4 hours it looks through your projects on its own and suggests what to do next. You can start one from scratch or from any of 24 ready-made templates. It works like a colleague who brings you a task list in the morning and hands the work out to assistants. But without its own VM, a bot runs only while your computer is on.
AI-processed from TG: @tips_ai (Tips AI / IT & AI); edited by Hamidun News
A typical AI coding assistant works on a question-and-answer model: until you ask something, it stays silent. Orca, an agentic development environment (ADE) for working with AI agents, has added a Bots tab to its sidebar, and the bots there behave differently. Every 4 hours, a bot checks your projects on its own and suggests next steps.
You can start one from scratch or from any of 24 ready-made templates. Imagine walking up to a colleague in the morning and finding that they have already gone through the correspondence, written down their conclusions and left a list of "what I would do today." For developers, the practical upshot is that some of the routine planning and task triage can be handed off to a bot that runs on your Claude subscription.
The details come from the author of the Telegram channel Tips AI, a longtime Orca user.
What a bot in Orca is and how to set one up
There are three ways to create a bot: from scratch, from a ready-made template (there are 24), or by importing from OpenClaw, Hermes and OpenBot. If you have already tried similar tools, you won't have to start from a blank slate.
A bot is simple by design: it is a folder of text files. SOUL.md describes its character and behavior rules, while MEMORY.md stores what the bot remembers about you and your projects. It runs on your Claude subscription or another model, and there is no separate bill for it. The upside is transparency: you can open the bot's behavior in a regular editor and tweak it, and you can put the folder under git.
There are not many connectors yet: 17 core ones, including Gmail, Notion and Cloudflare. If that isn't enough, you can connect a Composio account and use its integrations.
Features worth noting
The main one is proactivity. Every 4 hours, the bot checks your projects and suggests next steps. Every night, it rereads recent chats and updates MEMORY.md, so the next day it remembers what you agreed on yesterday.
A few more details from the review:
- skills can be linked from other projects on your computer, with no manual copying;
- when you connect Slack, Linear and similar services, you can allow or block specific people from calling your bot;
- bots can message each other, and several bots can be added to a single group chat.
Access control matters here. A bot that any member of a work Slack can call becomes an entry point into your projects and your subscription, so being able to choose who can invoke it looks like a sensible safeguard.
Worker orchestration and the main limitation
The most useful role for a bot right now is orchestrator. You give it a task, and it hands the work out to separate workers. Each worker runs in its own git worktree, an isolated working copy of the repository, so parallel tasks don't step on each other's files. Workers appear in the sidebar under the bot, and any thread opens in its own tab. Inside are an editor, a terminal, a browser, computer use and pull request review.
There is also a limitation, which the author states plainly. Without its own virtual machine (VM), a bot runs only while your computer is running. Close the laptop, and neither the nightly memory update nor the checks every 4 hours will happen. So for now bots have essentially one reliable use: managing other sessions while you are at the machine. An around-the-clock "on-call" assistant is not possible without your own VM.
What to do in practice
If you already work in Orca, try the Bots tab on one small project: create a bot from a template, see which steps it suggests and judge how useful the nightly memory is. If you are building your own agent system, you can borrow the approach itself: behavior and memory live in plain markdown files, and parallel tasks are isolated through git worktree. Both ideas work without Orca.
The post's author asks readers what they use, or plan to use, Orca Bots for, and it is a question worth asking yourself before you spin up a dozen bots with no clear job. Separately, he notes that, according to Thibault, the Dots feature has started rolling out to a small share of Plus users, but apparently only in the US.
- Nov 20, 2026Daydream Launches an AI Stylist That Buys Clothes Itself From 10,000 Brands
- Nov 20, 2026Developer Releases 30 SEO Skills for Claude Code and Cursor, Plus MCP Servers
- Nov 20, 2026Support team builds its own report with AI agents in 17 hours on a $20 subscription
- Nov 20, 2026Model ML Raises $250 Million for Banker-Focused AI Agents at a $1.5 Billion Valuation
Key Questions Answered
What a bot in Orca is and how to set one up?
There are three ways to create a bot: from scratch, from a ready-made template (there are 24), or by importing from OpenClaw, Hermes and OpenBot. If you have already tried similar tools, you won't have to start from a blank slate. A bot is simple by design: it is a folder of text files. SOUL.md describes its character and behavior rules, while MEMORY.md stores what the bot remembers about you and your projects. It runs on your Claude subscription or another model, and there is no separate bill for it.
What are the key details on Features worth noting?
The main one is proactivity. Every 4 hours, the bot checks your projects and suggests next steps. Every night, it rereads recent chats and updates MEMORY.md, so the next day it remembers what you agreed on yesterday. A few more details from the review: skills can be linked from other projects on your computer, with no manual copying; when you connect Slack, Linear and similar services, you can allow or block specific people from calling your bot; bots can message each other, and several bots can…
What are the key details on Worker orchestration and the main limitation?
The most useful role for a bot right now is orchestrator. You give it a task, and it hands the work out to separate workers. Each worker runs in its own git worktree, an isolated working copy of the repository, so parallel tasks don't step on each other's files. Workers appear in the sidebar under the bot, and any thread opens in its own tab. Inside are an editor, a terminal, a browser, computer use and pull request review. There is also a limitation, which the author states plainly.
Test Frontier LLMs & Autonomous Agents in One Workspace
Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.