AI Agent Memory Layer

Make your agents
remember everything.

Stop letting AI agents forget. Recall gives them persistent, searchable memory — so they learn from every interaction and never re-ask the same question.

3 lines to persistent memory
// Store a memory
await fetch("/api/recall/memories", {
  method: "POST",
  body: JSON.stringify({
    agentId: "support-bot",
    contextId: "customer-42",
    content: "User prefers email over phone. Had billing issue on March 3.",
    metadata: { source: "chat", importance: "high" }
  })
});

// Recall relevant memories
const res = await fetch("/api/recall/recall", {
  method: "POST",
  body: JSON.stringify({
    agentId: "support-bot",
    contextId: "customer-42",
    query: "What issues has this customer had?",
    limit: 5
  })
});

// Returns ranked memories with scores
// [
//   { content: "User prefers email...", score: 0.94 },
//   { content: "Billing issue resolved...", score: 0.87 }
// ]
The problem

AI agents have
amnesia by default.

⊘

Session amnesia

Every conversation starts from zero. Your agent re-asks questions it already knows the answers to.

⊞

Vector DB ≠ memory

Dumps create noise. Dumping everything into a vector DB returns irrelevant results and bloats costs.

⊘

No cross-session learning

Agent A can't learn from Agent B's interactions. No shared institutional knowledge.

⊘

Memory grows unbounded

Without summarization and decay, old memories crowd out recent ones and costs spiral.

What Recall does

Memory that actually works.

◎

Semantic search

Recall memories by meaning, not keywords. 'billing issue' finds 'invoice payment problem'.

⟐

Context grouping

Organize memories by customer, project, or session. Each context is an isolated knowledge base.

⚡

Automatic decay

Access frequency and recency influence ranking. Old, unused memories fade. Important ones persist.

✎

Summarization

Compress 100 memories into a concise summary. Cut token costs by 90% while preserving key facts.

◈

Importance scoring

Mark memories as critical, normal, or ephemeral. Critical memories never decay.

⊞

Multi-agent sharing

Agents can share a context pool. One agent learns from another's interactions.

Use cases

One API. Every agent type.

▸

Support bot

Remembers every customer's history, preferences, and open issues across sessions.

▸

Sales agent

Knows every prior conversation, objection, and deal stage — never re-asks.

▸

Coding assistant

Recalls your project conventions, past decisions, and codebase patterns.

▸

Research agent

Builds cumulative knowledge across research sessions instead of starting fresh.

How it works

3 lines of code.
Persistent memory forever.

STEP 01

Install the SDK

npm install recall-ai — or just use our REST API directly.

STEP 02

Store memories

agent.remember('User prefers dark mode') — one line, done.

STEP 03

Recall when needed

agent.recall('user preferences') — ranked by relevance and freshness.

Your agents deserve
to remember.

Start giving your agents memory today. Free tier includes 10,000 memories and unlimited search.

Get started free