Your infrastructure.
Your Stash.
Start with the environment your agents need. Plan local processing, integrations, access boundaries, and operations around your requirements.
Choose the environment.
Define the scope.
Cloud Managed
Discuss hosting, data handling, support access, and operational responsibilities for a managed environment.
- Hosting requirements
- Data handling boundaries
- Operational responsibilities
Private VPC
Plan a deployment within your private cloud environment. Define networking, access, integrations, and the operating model.
- Private network requirements
- Access and integration boundaries
- Operations planning
On-Premise
Run Recall Racoon on infrastructure you control. Configure local processing, memory storage, monitoring, and backups for your environment.
- Local infrastructure
- Configured local models
- Memory and backup operations
Air-Gapped
Define offline requirements before deployment. Review model availability, integrations, updates, and every dependency on external connectivity.
- Offline operation requirements
- Model and dependency planning
- Update and support procedures
Define the
boundaries.
| Area | What to establish |
|---|---|
| Infrastructure | Where the platform and storage will run. |
| Processing | Which local models and optional external integrations are configured. |
| Access | How agents and people connect within the intended scope. |
| Operations | Who handles monitoring, maintenance, and backups. |
| Recovery | Backup retention, offsite requirements, and restore testing. |
| Connectivity | Which workflows require external connections. |
What you need
to know
Local Processing
Use configured local models for embeddings and generation. Review the complete data flow for connected agents and integrations.
Monitoring
A built-in watchdog checks the full stack every five minutes and emails you the moment anything goes down or crashes — and again on recovery. The daily status report delivers a full health digest every morning. The dashboard reports true service state, so a green light means green.
Memory Operations
Manage stored knowledge through Mission Control, MCP, or REST. Watch the Live Activity feed as memories are saved, updated, merged, or archived.
Backup & Recovery
Backup tooling dumps the database and vector store in one step, and a pull script copies backups off the host. Disaster recovery is a documented procedure, not a hope.
Deployment Scope
Define the environment, integrations, and responsibilities before implementation. Support and service terms belong in the agreed deployment scope.
Growth Planning
Automatic six-month archiving moves untouched memories out of search into portable archives, keeping retrieval fast as the Stash scales. Tombstones preserve the record; re-import restores anything.
Embeddings and generation
on your terms.
Recall Racoon is local-first by default. You can run memory processing on your own infrastructure and use configured local models for both embeddings and generation. This gives you a deployment path that does not require an external AI API for embeddings or generation, keeping the core memory workflow within the boundaries you control.
The complete data flow is still defined by what you connect. Connected agents, optional integrations, backups, and support arrangements all determine where information travels. Reviewing these connections up front keeps the deployment predictable and scoped to what your teams actually need.
Know what connects
and where.
Access is scoped by namespace and project. Agents and people connect within the intended scope, and retrieval stays within that scope. Input validation establishes the context for each operation before content enters the memory workflow, while rate limiting controls request activity through configured limits.
Every deployment has a connectivity picture worth defining early: which workflows require external connections, which local models are configured, and who handles the ongoing operations. Laying this out before implementation keeps the deployment honest about what is self-contained and what is not.
Plan for continuity,
not surprise.
Memory backups can be scheduled with configurable retention, so your recovery window matches your operational needs. Optional offsite mirroring adds a second copy outside the primary environment, and restore testing verifies that recovery actually works before you need it.
Recovery planning is part of the operating model rather than an afterthought. Defining backup retention, offsite requirements, and restore testing as part of the deployment scope means the team responsible for operations has a clear, agreed path back when something goes wrong.
See health. Plan
for scale.
Service health and usage are visible through automated checks and daily status reports, so operators can review memory operations and identify issues that need attention. Mission Control provides a view of stored knowledge, session state, and memory activity.
As usage grows, review memory volume, retrieval activity, and processing requirements to match capacity to the workload. Growth planning keeps a growing Stash from becoming an unmanaged one, and keeps the deployment aligned with the retrieval patterns your agents actually produce.
Cloud, private, on-premise,
or air-gapped.
Deployment options span cloud managed, private VPC, on-premise, and air-gapped environments. A managed deployment lets you discuss hosting, data handling, support access, and operational responsibilities. A private VPC places the deployment inside your own cloud environment, where networking, access, and integrations follow your boundaries.
On-premise runs Recall Racoon on infrastructure you control, with configured local models, local memory storage, and your own monitoring and backup operations. Air-gapped deployments define offline requirements before anything ships: model availability, integration planning, update procedures, and every dependency on external connectivity is reviewed up front.
Who runs what,
agreed up front.
A clear deployment defines the environment, the integrations, and the operating responsibilities before implementation. Support and service terms belong in the agreed deployment scope, and the teams who handle monitoring, maintenance, and backups are identified early rather than discovered later.
The same clarity applies to the data path. Understanding which workflows require external connections, which local models are configured, and how credentials are handled keeps the deployment predictable. This is the difference between a memory platform that simply runs and one that your teams can operate with confidence over time.
Ready to deploy your own Stash?
Tell us about your agents, your infrastructure, and the knowledge they need to retain. We'll walk through the right deployment for your workflow.