The agent that never forgets.
And never waits to be asked.
Persistent support agents with customer context, follow-up memory, and safe access to the systems your team already uses.
Live workspace
Open workspace →Read-only preview of the Support workspace running on Lobu.
Architecture
Every agent runs in a sandboxed container with no outbound network by default — on your infrastructure, not ours.
- Link users across platforms with single sign-on
- Approval flows, rich cards, buttons, and more
- Workers never see secrets
- HTTP proxy with domain allowlist
- MCP proxy with per-user OAuth
- BYO provider keys (Anthropic etc.)
- One sandbox per user and channel
- Kata Containers / Firecracker microVMs / gVisor on GCP
- virtualized bash for scaling beyond 1000 users
- No direct internet access (internal network)
- Nix reproducible environments
- OpenTelemetry for observability
Lobu in memory benchmarks
Apples-to-apples comparison on public memory datasets. The model is GLM 5.1 via z.ai.
LongMemEval (oracle-50)
Single-session knowledge retention.
| System | Overall | Answer | Retrieval | Latency |
|---|---|---|---|---|
| Lobu | 87.1% | 78.0% | 100.0% | 237ms |
| Supermemory | 69.1% | 56.0% | 96.6% | 702ms |
| Mem0 | 65.7% | 54.0% | 85.3% | 753ms |
LoCoMo-50
Multi-session conversational memory.
| System | Overall | Answer | Retrieval | Latency |
|---|---|---|---|---|
| Lobu | 57.8% | 38.0% | 79.5% | 121ms |
| Mem0 | 41.5% | 28.0% | 66.9% | 606ms |
| Supermemory | 23.2% | 14.0% | 36.5% | 532ms |
Latest blog posts
Filesystem vs Database for Agent Memory
Agents need a workspace to think in and a warehouse to remember in. The filesystem is for ephemeral work. The memory layer is for durable organizational knowledge.
Add OpenClaw to your Next.js app in 10 minutes
Embed OpenClaw agents in a Next.js project with per-user sandboxing, MCP tools, and network isolation — no Docker required.
MCP Is Overengineered, Skills Are Too Primitive
MCP HTTP is great for external services. MCP stdio is redundant. And most skill systems are just prompt text with no reproducibility. Here's what we built instead.
Two ways to see it run.
Click through a live workspace, or book 20 minutes with the founder.