E
Engram

Your agent forgets.
Then it breaks.

Persistent memory API for autonomous agents. Store facts, retrieve context, score accuracy, and detect when your agent's knowledge is drifting before it costs you customers.

Two lines to integrate. Your agent gets smarter every retrieval — and tells you when it's getting dumber.

Free tier: 1 agent, 10K facts. No credit card required.

Store → Retrieve → Score → Detect Drift
# Store a fact
curl -X POST https://engram.cipherbuilds.ai/api/facts \
  -H "Authorization: Bearer eng_your_key" \
  -d '{"content": "Client prefers async, no calls", "tags": ["client"]}'

# Check memory health
curl https://engram.cipherbuilds.ai/api/facts/drift \
  -H "Authorization: Bearer eng_your_key"

# → { "drift_score": 73, "drift_status": "drifting", 
#     "summary": { "stale": 12, "low_confidence": 5 } }

The week-two problem

Day 1, your agent is sharp. It remembers client preferences, knows the codebase, nails the context. By week two, it's hallucinating old facts, missing updates, and making decisions on stale data.

This isn't a model problem. It's a memory problem. Context windows overflow. Session restarts lose state. Markdown files don't scale. And nobody notices the drift until a customer complains.

Felix, the most profitable autonomous agent online, paused his highest-margin service ($2K/setup) because memory degradation killed client trust. A $12K revenue stream — gone.

How it works

📥

Store

Your agent stores facts as it learns. Client preferences, decisions, context from conversations. Tagged, timestamped, scored.

🔍

Retrieve

Before any action, query for relevant context. Full-text + semantic search. Hot facts surface first, cold facts sink.

📊

Score

After acting, tell Engram if the context helped. Useful facts get promoted to hot tier. Misleading ones get demoted. Memory improves with every retrieval.

🔬

Detect Drift

Only Engram does this. Built-in drift detection monitors your agent's memory health. Stale facts, never-accessed data, decaying confidence — caught before they cause failures. Auto-decay keeps memory honest.

Drift detection: why it matters

Your agent's knowledge decays as reality changes. Endpoints move. Team members leave. Pricing updates. The agent doesn't know what it doesn't know.

Engram tracks every fact's freshness: when it was stored, last accessed, how often it was helpful, and whether it's been validated recently.

One API call tells you your agent's memory health score. Facts that go dormant decay automatically. Stale knowledge gets flagged before it causes bad decisions.

Drift Detection Response
{
  "drift_score": 73,
  "drift_status": "drifting",
  "summary": {
    "total_facts": 847,
    "drifting_facts": 134,
    "never_accessed": 41,
    "stale": 67,
    "low_confidence": 26
  },
  "facts": [
    {
      "content": "Deploy endpoint: api.old...",
      "drift_reason": "stale",
      "days_since_access": 18.3,
      "score": 0.21
    }
  ]
}

How Engram compares

FeatureEngramMem0ZepLetta
Store & retrieve
Retrieval scoringPartial
Hot/warm/cold tiers
Drift detection
Auto-decay
Fact maintenance API

Quickstart (2 minutes)

1. Create your agent

curl -X POST https://engram.cipherbuilds.ai/api/agents \
  -H "Content-Type: application/json" \
  -d '{"name": "my-agent"}'

# → { "api_key": "eng_abc123...", "tier": "free", "fact_limit": 10000 }

2. Store facts as your agent learns

curl -X POST https://engram.cipherbuilds.ai/api/facts \
  -H "Authorization: Bearer eng_your_key" \
  -H "Content-Type: application/json" \
  -d '{"content": "Client timezone is PST, prefers async", "tags": ["client", "prefs"]}'

3. Retrieve before acting

curl "https://engram.cipherbuilds.ai/api/facts/search?q=client+preferences" \
  -H "Authorization: Bearer eng_your_key"

4. Score what worked

curl -X POST https://engram.cipherbuilds.ai/api/facts/score \
  -H "Authorization: Bearer eng_your_key" \
  -d '{"fact_id": "abc123", "helpful": true}'

5. Monitor memory health

# Check drift status
curl https://engram.cipherbuilds.ai/api/facts/drift \
  -H "Authorization: Bearer eng_your_key"

# Run decay cycle (call weekly)
curl -X POST https://engram.cipherbuilds.ai/api/decay \
  -H "Authorization: Bearer eng_your_key"

# Re-validate facts you've confirmed are still accurate
curl -X POST https://engram.cipherbuilds.ai/api/facts/drift \
  -H "Authorization: Bearer eng_your_key" \
  -d '{"fact_ids": ["abc123", "def456"], "action": "validate"}'

API Reference

POST/api/agentsCreate agent, get API key
POST/api/factsStore a fact
GET/api/factsList facts (filter by tier, tag)
GET/api/facts/search?q=Search facts by text
POST/api/facts/scoreScore a retrieval (helpful/not)
GET/api/facts/driftCheck memory drift status
POST/api/facts/driftValidate, demote, or purge facts
POST/api/decayRun time-based score decay
GET/api/statsAgent memory stats & analytics
DELETE/api/facts?id=Delete a fact

Pricing

Start free. Upgrade when your agent needs more memory. Pay with card or crypto.

Free

$0/mo
  • ✓ 1 agent
  • ✓ 10,000 facts
  • ✓ Full-text search
  • ✓ Retrieval scoring
  • ✓ Drift detection
  • ✓ Auto-decay
  • ✓ Dashboard
Get Started
Most popular

Pro

$29/mo
or $290/yr with crypto (save ~10%)
  • ✓ 5 agents
  • ✓ 100,000 facts
  • ✓ Semantic search
  • ✓ Retrieval scoring + analytics
  • ✓ Drift detection + alerts
  • ✓ Auto-decay
  • ✓ Priority support

Team

$99/mo
or $990/yr with crypto (save ~17%)
  • ✓ Unlimited agents
  • ✓ 1M facts
  • ✓ Shared memory spaces
  • ✓ Cross-agent memory
  • ✓ Drift detection + webhooks
  • ✓ Auto-decay + custom rules
  • ✓ Dedicated support
💎

Pay with Crypto

Pay with USDC on Base. Per-transaction wallet, instant confirmation, auto-receipt.

Secure per-transaction wallets. Auto-sweep to treasury. Invoice generated on confirmation.

Stop your agent from forgetting

Free tier gets you started in 2 minutes. No credit card. No SDK. Just curl.

Get Your API Key — Free