Today you use tools. Soon, you'll describe outcomes and your butler uses the tools.
core-demo-github.mp4
Every AI agent you use today is smart. And every single one forgets you the moment the conversation ends.
We think that's broken. We think your AI should know you deeply, persistently, across every tool and every conversation. And once it knows you, it shouldn't sit there waiting for you to type. It should act.
Not recklessly. Intelligently. Sometimes it suggests and waits for your approval. Sometimes it just handles work because it's learned your patterns over weeks and months.
That's CORE. An AI butler that learns who you are, connects to your apps, and does the work before you ask. Open source, self-hostable, fully yours.
Your tasks are everywhere. A bug comes in on GitHub. A feature request lands in Linear. An action item is buried in a Slack thread. Another one hiding in an email you haven't read. Right now, you're the one tracking all of this in your head. CORE watches all of them and funnels everything into one place — tasks created automatically, nothing falls through cracks.
You're the glue between your tools and your coding agents. Before you can even start a Claude Code session, you spend 15 minutes gathering context — find the issue, read the thread, check related PRs, look at the error logs, then paste it all in. CORE does that job for you. Tell it "fix the checkout bug from issue #312" — CORE pulls the issue details from GitHub, searches memory for past decisions about that service, checks Sentry for the error trace, and spins up a Claude Code session with all that context already loaded. You come back to a PR.
You're running multiple agent sessions and losing track. Three Claude Code terminals, each working on something different, and you're context-switching between them trying to remember what's where. CORE is the orchestrator — kick off sessions from WhatsApp, Slack, or the dashboard. Each one gets the right context from memory and toolkit. CORE tracks what's running, reports back when things are done, and you stay focused.
At 2 AM, a Sentry alert fires. CORE has seen you triage these before — you always check the deployment diff first, then the error trace, then assign to whoever last touched that file. It does exactly that. Spins up a Claude Code session, investigates, creates the issue, assigns the right engineer. You wake up to: "Handled. Issue #847, assigned to Harshith." You didn't ask for any of this.
The difference? Every interaction makes CORE smarter. It learns your patterns, your codebase preferences, your team structure. It knows when to ask you for approval and when to just handle it — because it's been watching and learning. That's what a butler with memory does that a tool without memory can't.
- ✅ You're tired of being the glue between your AI tools — re-explaining context, copying information, manually kicking off workflows
- ✅ You want an AI that actually remembers your preferences, decisions, and relationships and gets better over time
- ✅ You want proactive automation, not just reactive chat — an AI that watches your apps and acts on what matters
- ✅ You want to say "fix the checkout bug" on WhatsApp and it comes back with a PR
- ✅ You care about owning your AI — self-hosted, open source, your data never leaves your infra
- ✅ You want one brain across WhatsApp, Slack, email, and your coding tools — not five separate contexts
| Without CORE | With CORE |
|---|---|
| ❌ You re-explain your project, stack, and preferences to every AI tool, every session. | ✅ CORE remembers everything — preferences, decisions, relationships, project context — and it gets sharper over time. |
| ❌ Your AI tools can take actions, but they can't decide which action to take for you. They don't know you. | ✅ CORE's temporal memory means it understands nuance: who this client is, how you handle Sentry alerts, what your morning priorities look like. |
| ❌ You are the bottleneck. Every workflow starts with you typing something. | ✅ CORE watches webhooks, polls your tools, and acts proactively — sends you a morning brief, triages alerts, drafts replies. |
| ❌ You have an idea at midnight. You have to find the repo, open a terminal, set up context, babysit the agent. | ✅ Message CORE on WhatsApp: "migrate the auth service to Postgres." Come back to a PR. |
| ❌ Your chat history is a graveyard. Decisions, preferences, and context die with every session. | ✅ Every conversation feeds a knowledge graph. Facts are classified (preference, decision, goal, directive) and connected over time. |
| ❌ Different AI tools, different contexts, no shared understanding of who you are. | ✅ One brain. WhatsApp, Slack, email, Claude Code, Cursor — same memory, same butler. |
| Not a chatbot. | Your butler has a name, a memory, and a job. It's not waiting for your prompt — it's watching your apps and acting on what matters. |
| Not a RAG wrapper. | Memory isn't "embed chunks and search." It's a temporal knowledge graph where facts are classified, connected, and updated over time. It knows when you decided something and why. |
| Not a copilot plugin. | Copilots assist when you're working. CORE works when you're not. It's proactive, not reactive. |
| Not a workflow builder. | No drag-and-drop. You talk to CORE in natural language. It figures out the workflow. |
Your tasks are scattered — GitHub issues, Linear tickets, email action items, Slack requests. CORE watches all of them and funnels everything into one place (Todoist, Google Tasks). A GitHub issue gets assigned to you? Task created. An email has an action item buried in paragraph three? Task created. No manual entry, no things falling through cracks.
Once your tasks are in one place, CORE helps you plan. It knows your calendar, your deadlines, your priorities. Ask it to plan your day and it builds a schedule from actual data — not a generic to-do list.
This is where it gets powerful. CORE doesn't just track tasks — it can do them. A user reports a bug over email → CORE alerts you, creates a Todoist task, and when you say "handle it" → it spins up a Claude Code session, investigates the issue, writes the fix, opens a PR. You come back to a summary and a link.
Tell CORE your rules once: "When a Sentry alert fires at night, investigate it." "When an email needs a response, draft one for me." CORE watches your tools 24/7, evaluates events against what it knows about you, and either acts or queues things for your morning review. You wake up to work already done — not a pile of notifications.
Every use case above gets better over time. The more you use CORE, the more it learns — your preferences, your patterns, your relationships. That's what the temporal knowledge graph is for. Not just storing information, but understanding you.
Open source. Self-hosted. No account required.
git clone https://github.com/RedPlanetHQ/core.git
cd core/hosting/docker
cp .env.example .env
# Add your model key (see below)
docker compose up -dOpen http://localhost:3000 → connect your first app → hand off your first task.
Pick your model (edit .env):
OPENAI_API_KEY=sk-... # OpenAI
ANTHROPIC_API_KEY=sk-ant-... # Anthropic Claude
OLLAMA_URL=http://localhost:11434 # Ollama — fully local, no cloud
OPENAI_BASE_URL=https://your-proxy.com # Any OpenAI-compatible endpointRequirements: Docker 20.10+, Docker Compose 2.20+, 4 vCPU / 8GB RAM
☁️ Prefer cloud? CORE cloud is in early access. Join the waitlist →
CORE achieves 88.24% average accuracy on the LoCoMo benchmark — single-hop, multi-hop, open-domain, and temporal reasoning.
Want to understand how CORE works under the hood?
- Memory — Temporal knowledge graph, fact classification, intent-driven retrieval
- Toolkit — 1000+ actions across 50+ apps via MCP
- CORE Agent — The orchestrator: triggers, memory, tools, sub-agents
- Gateway — WhatsApp, Slack, Telegram, email, web, API
- Skills & Triggers — Scheduled automations and event-driven workflows
- API Reference — REST API and endpoints
- Self-hosting — Full deployment guide
- Changelog — What's shipped
CASA Tier 2 Certified · TLS 1.3 in transit · AES-256 at rest · Your data is never used for model training · Self-host for full isolation
Security Policy → · Vulnerabilities: harshith@poozle.dev
We're building the future of personal AI in the open. Come build with us.
- Discord — questions, ideas, show-and-tell
- CONTRIBUTING.md — how to set up and send a PR
good-first-issue— start here
Own your AI, not just use one.