Memory API · $19/month

Your AI agents forget everything.Fix it in 3 lines.

Add persistent memory, semantic recall, and context compression to any AI agent — LangChain, AutoGen, CrewAI, or raw API calls. One SDK. One minute.

Start for $19/moSee how it works

No hallucinations. No stateless agents. Just memory.

agent.py
from synap import SynapClient
client = SynapClient(api_key="syn_...")
client.remember(agent_id="agent-1", content="User prefers bullet points")
memories = client.recall(agent_id="agent-1", query="user preferences")
● memory storedrecalled: 5 resultslatency: 7ms
Works with LangChainAutoGenCrewAIany LLM< 10ms p99Zero config

Capabilities

Memory infrastructure for the agentic era

Everything your agents need to remember, recall, and compress — served at infrastructure scale.

Persistent Memory

Agents remember users, preferences, and past interactions across every session — no more stateless conversations or repeated onboarding.

Semantic Recall

Retrieve the most relevant memories by meaning, not keywords. Vector embeddings + composite relevance scoring under 10ms p99.

Context Compression

Compress long conversation histories into dense, queryable representations. Cut token usage by up to 80% without losing signal.

Plug-and-Play API

Three methods. One SDK. Works with OpenAI, Anthropic, Mistral, and any LLM. No new infra, no new databases — just import and go.

Integration

Three methods. That's the whole API.

No graph databases to configure. No vector index tuning. No ops overhead.

01

Store a memory

Write anything your agent learns — user preferences, facts, conversation highlights. Assign an importance score to control retention.

client.remember(agent_id="agent-1", content="User prefers bullet points", importance=0.9)
02

Recall relevant context

Query by meaning. Synap returns the top-k most relevant memories ranked by semantic similarity + recency + importance — ready to inject into your next LLM prompt.

memories = client.recall(agent_id="agent-1", query="user preferences", top_k=5)
03

Compress long histories

Before a long conversation hits your context limit, compress it into a dense summary stored as a memory. Resume sessions with full context, zero token waste.

summary = client.compress(agent_id="agent-1", context=messages, max_tokens=512)
Get API access — $19/mo

Pricing

One plan. Full API.

No free tier complexity. No usage-based surprises. Just the full API for a flat $19/month.

Pro

$19/month

Cancel anytime. No contracts.

  • 1M memory operations / month
  • Unlimited agents
  • 365-day memory retention
  • Python SDK included
  • Semantic recall + context compression
  • Priority support + SLAs
Start for $19/month

Secure checkout via Stripe · Starter API key delivered instantly

FAQ

Common questions

What is Synap Memory API?

Synap is a memory-as-a-service API for AI agents and LLMs. It provides three primitives: persistent storage of memories (remember), semantic retrieval by meaning (recall), and context compression to reduce token usage (compress). You call the API; Synap handles the embeddings, vector search, decay scoring, and storage.

How is this different from a vector database?

A vector database is a search index — you manage the schema, the embeddings model, the query logic, and the infra. Synap Memory API is a complete memory OS. It automatically handles embedding generation, relevance scoring (recency + importance + semantic similarity), memory decay, deduplication, and context-aware compression. You call .remember() and .recall(), not SQL.

Does it work with my existing LLM framework?

Yes. Synap is framework-agnostic. It works with LangChain, AutoGen, CrewAI, LlamaIndex, Semantic Kernel, raw OpenAI/Anthropic API calls, and anything else. You call the Synap SDK before or after your LLM call — no framework-level integration required.

What counts as a memory operation?

One operation = one API call. A remember() call is 1 op. A recall() call is 1 op regardless of how many results are returned. A compress() call is 1 op. At 1M ops/month on the Pro plan, most production agents run well within budget — a high-traffic agent handling 10K conversations/day uses roughly 600K ops/month.

Can I cancel anytime?

Yes. Cancel any time via your billing portal — no contracts, no cancellation fees. Your access continues until the end of the current billing period.

Is there a free trial?

There is no free tier — Synap Memory API is a paid product at $19/month. The first month is yours to evaluate: if the API doesn't work for your use case within 30 days, email us for a full refund.

Ready to give your agents a memory?

Get started — $19/month

30-day money-back guarantee · Cancel anytime