Corellis
Corellis is a self-hosted system that turns a single OpenClaw assistant into a self-managing AI workforce of 'lobsters' (individual AI assistants) that coordinate tasks, share memory, and improve over time. It solves the problem of scaling AI assistance beyond one person by providing persistent team members with collective memory and fleet learning. It is for developers and teams who want to deploy their own multi-agent AI organization on their own infrastructure.
✨ Key features
- Teamind: semantic search across all team Slack messages
- Self-improving lobsters with fleet-wide learning
- GoalOps: decompose high-level goals into coordinated sub-goals
- Unified task board with sprint planning and breakdown
- 30-second spawning of new lobsters via Slack app creation
- 17+ built-in skills including deep research and weekly reports
🎯 Use cases
- Automate marketing campaigns by giving a high-level goal
- Coordinate cross-team projects with autonomous task decomposition
- Maintain a searchable collective memory of all team decisions
- Run a fleet of AI assistants for ops, finance, and research
- Implement a coding agent workflow with confidence-based routing
📦 Installation
🧰 Requirements: Requires OpenClaw on the host machine, Docker + Docker Compose, a Slack workspace with bot apps (one per lobster), and 2GB RAM per lobster (3GB with Claude Code).
git clone https://github.com/CorellisOrg/corellis.git
cd corellis
cp .env.example .env # add your LLM API key
docker compose up -d # launch your first lobster
Note: Each lobster needs a Slack bot app. Run
./scripts/create-slack-app.sh <name>to create one automatically, or see Slack Bot Setup for manual setup.
🚀 Usage
# After installation, tell your controller to spawn a lobster:
# "spawn a lobster for alice"
# The controller creates the Slack app, handles OAuth, and launches the container.
# You click Allow and paste one token.
For a full walkthrough, see the Tutorial: Set Up a 3-Person Team in 30 Minutes.
❓ FAQ
How does Corellis differ from CrewAI or AutoGen?
CrewAI and AutoGen orchestrate tasks, while Corellis orchestrates an organization with persistent lobsters that have memory, identity, and relationships.
What are the hardware requirements?
Each lobster requires 2GB RAM (3GB with Claude Code). The production fleet runs 28 lobsters on a single server with 64GB RAM and 16 vCPU.
Can I use Corellis with my existing task board?
Yes, the task management is backend-agnostic and works with Notion, Linear, GitHub Projects, or plain markdown.
How do lobsters learn from mistakes?
Lobsters detect when they are corrected, record the lesson, and permanently improve. Validated patterns are promoted fleet-wide so one lobster's mistake becomes every lobster's knowledge.
📊 Repository
🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.