Turn data into shared, structured memory
Owletto gives all your agents the same durable graph: connectors, recall, and managed auth without leaking credentials to the runtime.
Two ways to give agents memory
Per-agent files keep memory local to one sandbox. Lobu Memory gives every agent a shared, auditable graph through MCP.
Each agent has its own filesystem
- no cross-agent recall
- audit is per-agent and manual
- tied to one sandbox / session
Agents share Lobu Memory through MCP
- one truth across agents
- dedup via entity model
- inspectable + correctable
How it works
Turn scattered prompts, tools, and application data into a shared context layer your agents can use everywhere.
Model the world
Represent incidents, services, deploys, and pull requests as first-class objects so on-call context survives after the thread scrolls away.
Connect sources
Turn live operational signals into structured incident memory.
A scheduled watcher polls the channels and tools your team already uses, then writes the relevant signals back to the right entity. The prompt is the filter — chatter that doesn't match never becomes memory.
watcher.poll(every: 30m)- Incident INC-4421 — checkout-v43 rollback in progress (Dan)
- Priya blocked on checkout cluster admin creds
- Caching layer PR pending merge by EOD (Jay)
- INC-4378 postmortem drafting (Nina)
Let users connect their data
Let teams bring the tools they already use, while keeping credentials outside the worker.
Let teams bring the tools they already use, while keeping credentials outside the worker.
| Account | Brought by | Access | Used for |
|---|---|---|---|
| GitHub / GitLab | User | OAuth | PRs, commits, diffs |
| Slack / Linear / Notion | User | OAuth | Notes, tickets, team context |
| PagerDuty / Datadog | User or admin | OAuth / token | Alerts and incident state |
| AWS / GCP / internal APIs | Org admin | Service account | Infra and deploy metadata |
| Incident history | Org | Import / sync | Memory bootstrap |
Reuse context across agents
The same incident memory powers operational agents wherever teams work.
The same incident memory powers operational agents wherever teams work.
Keep it fresh
Watchers pull in new alerts, deploy state, and merged fixes so the runtime sees the latest impact and rollback options.
A scheduled watcher keeps this memory current as new source changes arrive.
{ incident_state, impacted_regions[], rollback_ready, blocking_prs[] }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.
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.
Introducing Lobu
From a Slack bot to multi-tenant OpenClaw infrastructure — the story behind Lobu.
Beats other memory systems on public benchmarks
Apples-to-apples comparison on public memory datasets. Same answerer (glm-5.1) and same questions.
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 |
Start building shared memory
Model the right entities, connect your sources, and keep long-term context available across every agent workflow.