Long-running engagements where context matters
Customer success, account management, research projects — work where today's question depends on what was learned six weeks ago.
Agent
OpenClaw + GBrain
OpenClaw with a brain that compounds.
OpenClaw + GBrain pairs the open-source workhorse agent with a persistent local brain — facts, decisions, files, and people the agent has encountered are remembered across sessions instead of re-explained every time. On MindsHub it deploys in one click: agent, brain, model router, and 43 pre-loaded skills, no setup.
What it is
What is OpenClaw + GBrain?
GBrain is a local memory layer for agents — a structured store of pages, facts, and entities that the agent reads from before answering and writes to as it works. Layered onto OpenClaw, it turns a stateless agent into one that gets sharper week over week: it remembers the customer you asked about last Tuesday, the schema decision you made last sprint, the playbook you reviewed last month.
On MindsHub, OpenClaw + GBrain ships as a 1-click deploy. Each user gets their own isolated instance with GBrain pre-installed (pinned to a verified upstream release), a 43-skill skillpack already scaffolded, PGLite as the local brain backend (no external database to wire up), and the MindsHub Model Router pre-configured. The full stack provisions in ~2–3 minutes and is ready to use without touching a terminal.
Especially good for
The work OpenClaw + GBrain is shaped to do.
Personal AI coworkers that learn your world
Your people, projects, decisions, and preferences — encoded once, reused forever. The agent stops re-asking questions you already answered.
Knowledge accumulation across teams
Standups, design reviews, postmortems — every conversation feeds the brain, so the agent can answer 'what did we decide about X?' months later.
Research and synthesis
Sources, claims, contradictions tracked over time. The agent can surface 'we saw the opposite finding in the March report' without you remembering it existed.
Founder / operator copilots
An agent that remembers your investors, your customers, your team's history — without you re-priming it every Monday morning.
Self-hosted vs MindsHub
What you build vs what you get.
OpenClaw + GBrain is open-source either way. The difference is the operational backbone around it.
| What you handle… | Self-hosted | On MindsHub |
|---|---|---|
| Secrets management | You manage keys, rotation, scoping | Built-in credentials vault |
| Model access | One provider relationship per LLM | Model Router with `latest:*` aliases |
| Persistent execution | Queues, workers, retries you build | Managed runtime, jobs survive restarts |
| Memory across runs | Your storage, your schema, your reset logic | Shared memory store, built in |
| Tool integrations | Per-tool plumbing, per-tool auth | Pre-built catalog, shared auth |
| Observability | You wire your own traces | Scratchpad + audit log on every run |
| Scaling | VMs, queues, autoscaling rules | Managed — no servers to mind |
Why on MindsHub
What you get when OpenClaw + GBrain runs here.
1-click deploy of the full stack
Agent, brain, model router, 43 skills — provisioned in ~2-3 minutes. The path from 'I want to try this' to a working agent is a single click, not a 2-hour setup.
Isolated per-user brain
Your brain lives on your instance. Other users' agents can't read or write to it. The data layer is private by construction, not by policy.
Pre-wired Model Router
The agent talks to a hosted LLM gateway out of the box — no OpenAI / Anthropic key required to start. Bring your own keys later if you want direct provider control.
Skillpack pre-scaffolded
GBrain ships with 43 skills (brain-ops protocols, ingestion patterns, query workflows) already installed and brain-first compliant. The agent uses them from session one.
Persistent execution and memory
Long-running tasks survive restarts. Yesterday's facts inform today's answers. The brain compounds across every session — that's the point.
No ops to run
No PGLite to set up, no embedding model to choose, no autopilot daemon to wire up. MindsHub manages the layer; you describe the work.
How it works
From signup to first run.
- 01
Deploy
One click. MindsHub provisions an isolated instance with OpenClaw + GBrain, the skillpack, the model router, and a per-instance brain.
- 02
Ingest
Point the agent at the things it should remember — documents, conversations, people, decisions. GBrain extracts and stores; every future session reads from the same memory.
- 03
Work
Hand off tasks the way you'd use OpenClaw, with one difference: the agent queries the brain first. Answers come back grounded in what you've taught it, not just general knowledge.
Other agents
Browse the rest of the platform.
FAQ
Frequently asked.
- What's the difference between OpenClaw and OpenClaw + GBrain?
- OpenClaw is the agent runtime — planning, tool use, code execution. OpenClaw + GBrain adds a persistent local memory layer (GBrain) so the agent's context survives across sessions. Plain OpenClaw answers in-session; OpenClaw + GBrain answers with everything it has ever learned. Same hosting tier on MindsHub, different shape of work.
- Is GBrain open source?
- Yes. GBrain is open-source (github.com/garrytan/gbrain). MindsHub deploys a verified upstream release pinned at bake time, so every instance runs bit-identical code.
- Where does the brain live? Is my data private?
- Each MindsHub instance gets its own isolated brain stored locally on the instance (PGLite, no shared database). Other users' agents have no read or write access to your brain. It's per-tenant by infrastructure, not by access-control policy.
- Do I need an OpenAI or Anthropic key to use it?
- No — the MindsHub Model Router is pre-wired for chat out of the box. For full GBrain features like embedding and the nightly dream cycle, you'll want to add your own OpenAI key (via gbrain config set openai_api_key sk-...) so the brain can index your content. We're working on routing embeddings through the platform too, which would remove this step.
- How is OpenClaw + GBrain priced on MindsHub?
- Same $9.95/month hosting as the other agents. Bring your own LLM keys or use the Router (5M tokens included). The brain itself runs on the local instance — no separate database cost. See /pricing.
- Can I run OpenClaw + GBrain locally instead?
- Yes — both OpenClaw and GBrain are open-source. The local install path is about two hours of setup for a technical user (compose the agent, install GBrain, configure embeddings, scaffold skills). MindsHub's 1-click is that path automated, plus the hosting infrastructure.