Memory
THE MEMORY STACK
What is Backboard memory?
Production-grade memory for apps that need to remember people, projects, and decisions over time—across channels, devices, and models.
Attach memory to workflows
Attach durable memory to users, teams, and workflows so context persists beyond a single session or chat history.
Store, organize, retrieve
Let Backboard store, organize, and retrieve what matters, surfacing the smallest sufficient context instead of replaying everything.
Portable across models
Call any model and have the right memories injected automatically through a stateful API that works across apps and LLMs.
Checkpoint branching
NEW
✓ Branch conversational threads without losing history
✓ Resume mid-flow after deploys and cold restarts
✓ Named checkpoints for safe experimentation
Retrieval fan-out
NEW
✓ Scale read traffic across hot memory slices
✓ Route queries to the healthiest index path
✓ Keep latency predictable under concurrent agents
PRODUCTS
Highest performance
across every use case
Everything you need to build production-grade AI — memory, state, RAG, and 17,000+ LLMs.
LIFECYCLE
Store · Route · Retrieve · Evolve
The same loop you use to ship reliable agents: persist state, trace what happened, evaluate retrieval quality, and ship improvements without schema drama.
Store
Ingest messages, tool outputs, and documents into a unified memory graph with tenancy and retention you control.
Route
Policy-aware routing sends each turn to the right model and the right memory slice—no manual wiring per session.
Retrieve
Structured traces show what was pulled and why—debug long contexts and flaky recall without guesswork.
Evolve
Ship new indexes and extractors behind API versions—backward compatible upgrades, zero cold migrations.
DEVELOPER-FIRST API