Remember
Everything
One memory that follows you across every AI. Save something in ChatGPT, recall it in Claude or Cursor — never summarized — stored word-for-word, with secrets and PII auto-redacted, searchable everywhere. The conversation is the knowledge base.
// Store a full conversation
await mcp.append_messages({
conversation_id: "conv_abc123",
messages: [
{ role: "user", content: "Deploy the API to prod" },
{ role: "assistant", content: "Running deploy..." },
{ role: "tool", tool_name: "deploy", content: "Success: v2.1.0" },
]
});
// Search later — semantic, instant
await mcp.search({ query: "when did we deploy v2?" });Built for agents that remember
Every existing memory product compresses your data. Engram stores the full transcript — because context is everything.
Verbatim Storage
Every message stored as full, unsummarized text — the complete transcript, not a lossy summary. Secrets and PII are auto-redacted before storage; everything else is preserved exactly.
Semantic Search
Find anything across thousands of conversations instantly. Vector search with automatic embedding at the edge.
MCP Native
Built on the Model Context Protocol. Connect any MCP-compatible agent — Claude, GPT, or your own — in minutes.
Edge-First
Powered by Cloudflare Workers, D1, and Vectorize. Sub-50ms latency globally. Scales to zero when idle.
Multi-Tenant
Organization-level isolation by default. API keys, tenant scoping, and audit trails built in from day one.
Memory Stays Clean
Memory is for content, not credentials. Recognized API keys, tokens, and PII are automatically detected and stripped from anything you save to memory, so a stored transcript doesn't casually carry a secret. For anything you need to keep on purpose, use the vault — never memory.
Zero-Knowledge Vault
Need to store a secret? The encrypted vault does exactly that. Values are AES-256-GCM encrypted on your machine before upload — the key never leaves your device, so the server can't read your secret values. Separate from memory, so your agent can use a secret without it ever entering searchable context.
Tool Call Tracking
Store tool invocations alongside messages. Replay agent workflows and debug with full context.
Engram vs. the rest
Other products extract and compress. Engram stores everything.
| Feature | Engram | Mem0 / Zep / etc. |
|---|---|---|
| Full verbatim storage | ||
| Semantic search | ||
| MCP native | ||
| Edge deployment (<50ms) | ||
| No message compression | ||
| Tool call tracking | ||
| Multi-tenant isolation | ||
| Conversation replay | ||
| Memory extraction |
Simple, transparent pricing
Start free. Scale when you need to. No surprise bills.
Free
For side projects and experimentation.
- 10,000 messages of memory
- Permanent memory — never expires
- Unlimited conversations
- 1 seat
- Semantic search
- MCP + REST API
- Conversation export
Pro
For developers building production agents.
- 1,000,000 messages of memory
- Permanent memory — never expires
- Unlimited conversations
- 1 seat
- Semantic search
- MCP + REST API
- Conversation export
Team
For teams that need shared memory.
- 1,000,000 messages of memory / seat, pooled
- Permanent memory — never expires
- Unlimited conversations
- Unlimited seats
- Semantic search
- MCP + REST API
- Conversation export
- Webhooks
- Usage dashboard
Enterprise
For organizations with scale and compliance needs.
- Unlimited memory
- Permanent memory — never expires
- Unlimited conversations
- Unlimited seats
- Unlimited API keys
- SSO / SAML
- SLA guarantee
- Custom data retention
- Security review & custom DPA
Give your agents perfect memory
Start storing conversations in under 5 minutes. Free tier included.