ON-PREMISES DEPLOYMENT

Your infrastructure.
Your data. Your control.

Deploy Recall on your own servers. Every document, every embedding, every query stays within your network. No cloud dependency, no external calls, no compromise on privilege.

ARCHITECTURE

HOW IT DEPLOYS.

Recall Server runs as a lightweight backend service. Clients connect from any machine on your network.

Your Server
Recall Server (.NET) Embedding Engine Document Store SQLite Database
ALL SERVICES RUNNING
LAN / VPN
Client Machines
Recall Desktop (macOS) Recall Desktop (Windows)
2 CLIENTS CONNECTED
ENTERPRISE CAPABILITIES

BUILT FOR FIRMS THAT TAKE
SECURITY SERIOUSLY.

Everything in Recall Desktop, plus the infrastructure controls your IT team expects.

FIRM-WIDE POWER

One server. Your entire firm.

Deploy Recall on your office server so every attorney can access the same research. It works on your existing hardware with one simple Docker setup.

DEPLOYMENT STATUS
docker compose up -d running
recall-server healthy
recall-embedder healthy
STAYS IN THE BUILDING

Zero external calls.

Your client files never touch the internet. Recall works entirely behind your office firewall, keeping your research truly private and protected.

AIR-GAPPED CAPABLE
SHARED ACCESS

Collaborative research.

Multiple attorneys on the same case. Share document folders, coordinate your research, and manage your firm's archives in one central place.

MULTI-USER CONCURRENT
YOUR LIBRARY, CENTRALIZED

Every filing, brief, and opinion in one place.

Store your firm's entire document library on your own server. Your IT team manages backups and security using the tools they already know.

LIBRARY STATISTICS
Active matters 47
Documents indexed 12,847
Total embeddings 1.2M vectors
FLEXIBLE

Built for your firm.

Recall adapts to your office setup. Your IT team chooses the models, storage, and performance tier that fits your specific needs.

COMPLIANCE

Meets your rules.

Keep data in your jurisdiction. No complex workarounds — just simple, local control over your firm's data and regulatory requirements.

SUPPORT

We help you deploy.

Our team works directly with your IT staff to plan, configure, and validate your on-premises installation. Ongoing support included.

SERVER SPECIFICATIONS

WHAT YOU NEED TO
GET STARTED.

Recall Server scales from single-rack deployments to full data center installations.

Standard

Small to mid-size firms (5–25 users)

  • CPU16+ cores (Xeon / EPYC)
  • RAM64 GB ECC
  • GPUNVIDIA A4000 (16 GB VRAM)
  • Storage1 TB NVMe SSD
  • OSUbuntu Server 22.04 LTS
  • RuntimeDocker + Compose

Max models at this tier

Embedding model ~400M – 600M params
Large language model Up to ~70B params (incl. MoE)

70B-class models on a single GPU. Fast concurrent inference for the entire firm.

Enterprise

Large organizations (100+ users)

  • CPU64+ cores (dual EPYC)
  • RAM512 GB+ ECC
  • GPUMulti-GPU (H100 80 GB x4–x8)
  • Storage10 TB+ NVMe RAID
  • OrchestrationKubernetes
  • NetworkingInfiniBand / NVLink

Max models at this tier

Frontier language model ~400B – 1T+ params (incl. MoE)
Any open-weight model no upper limit

Frontier-class models at full precision. Multi-GPU tensor parallelism across H100 clusters with NVLink.

READY TO DEPLOY ON YOUR TERMS?

Talk to our team about getting Recall running on your infrastructure.

FREE EVALUATION FOR QUALIFIED FIRMS. VOLUME LICENSING AVAILABLE.