GIR MODE ACTIVATED
Open Source · Self-Hosted · Rust Native

Your agents have amnesia.

Engram is a cognitive memory system that learns what matters, forgets what doesn't, and builds knowledge graphs that grow smarter over time. No external APIs. No vendor lock-in.

RUST_NATIVE// ONNX_EMBEDDINGS// FSRS_6_DECAY// KNOWLEDGE_GRAPHS// HYBRID_SEARCH// AT_REST_ENCRYPTION// MULTI_TENANT// AGENT_FORGE// MNEMONIC_SIDECAR// TOOL_COMPRESSION// PER_AGENT_KEYS// CLAUDE_HOOKS// EIDOLON// CREDD// MILLION_PLUS_SCALE// HOPFIELD_NETWORK// DREAM_CONSOLIDATION// CROSS_ENCODER_RERANK// RUST_NATIVE// ONNX_EMBEDDINGS// FSRS_6_DECAY// KNOWLEDGE_GRAPHS// HYBRID_SEARCH// AT_REST_ENCRYPTION// MULTI_TENANT// AGENT_FORGE// MNEMONIC_SIDECAR// TOOL_COMPRESSION// PER_AGENT_KEYS// CLAUDE_HOOKS// EIDOLON// CREDD// MILLION_PLUS_SCALE// HOPFIELD_NETWORK// DREAM_CONSOLIDATION// CROSS_ENCODER_RERANK//

Everything agents need
to remember.

Modular capabilities that compose into a full cognitive memory layer. Built in Rust for speed, safety, and zero runtime dependencies.

📈

FSRS-6 Decay

Power-law forgetting that models human memory. Memories fade unless reinforced.

🔗

Knowledge Graph

Entities, communities, PageRank. See how memories connect and which matter most.

🔍

Hybrid Search

Semantic vectors + BM25 keywords + graph context. Cross-encoder reranking.

🔐

At-Rest Encryption

SQLCipher database encryption. Password, hardware key, or TPM-backed. Argon2id KDF.

👤

Personality

Learned preferences, style guides, and user-specific behaviors that persist.

🕐

Time Travel

View memory state at any point in time. See what the agent knew and when.

🤖

Multi-Agent

Shared memory pools, agent namespaces, cross-agent knowledge transfer.

📌

Decomposition

Break complex memories into atomic facts. Each scored and decays independently.

🎯

Budget Context

Token-aware assembly. Fill context windows optimally for any model.

📡

Eidolon

Activity tracking and observability. Session logs, task progress, agent heartbeats.

🔑

credd

Secure credential daemon. Agent keys, secret resolution, proxy injection. Never hardcode again.

🧠

Hopfield Network

Modern associative memory (Ramsauer 2020). Pattern completion with exponential capacity.

🌙

Dream Consolidation

Sleep-like memory processing. Discover associations, merge similar, prune weak, replay important.

🔗

Causal Chains

Track cause/effect relationships. Build reasoning chains across memories with confidence scoring.

🎯

Cross-Encoder Rerank

IBM Granite ONNX model. Semantic precision on top-K results after initial retrieval.

🛠

Tool Grounding

Secure tool execution. MCP, Shell, Web, GUI backends with sandboxing and security policies.

📈

Million+ Scale

LanceDB vector index handles 1M+ memories per user. Single-user or production multi-tenant.

🛠

Agent-Forge

Structured reasoning protocol. spec_task, log_hypothesis, verify, challenge_code. Quality workflows baked in.

💭

Mnemonic Sidecar

Session companion. Observes tool use, auto-flushes to memory, compresses outputs, recalls context.

🔒

Per-Agent Keys

Every agent and model gets its own encrypted API key. Full audit trail of who stored what.

Claude Hooks

Drop-in hooks for Claude Code. Auto context loading, session tracking, memory storage. Zero manual prompting.

A day with Engram

What it actually looks like to use persistent memory.

Session Starts

Pick up where you left off

Your agent pulls context from Engram. It knows the project state, your preferences, what you decided last time, and what's still unresolved.

While Working

Decisions and discoveries stored automatically

New architecture choices, deployment configs, bug resolutions -- all stored with importance scoring and auto-linked to related memories.

Conflict Detected

"You said X last week but now Y -- which is correct?"

Engram catches contradictions between new information and existing knowledge. Your agent surfaces them instead of silently overwriting history.

Session Ends

Memories consolidate, weak ones fade

Important memories grow stronger. Irrelevant details lose retrieval strength. The knowledge graph reorganizes around what actually matters.

Next Session

Your agent remembers everything that mattered

Not a transcript dump. Not a vector search over flat files. Weighted, prioritized, personality-aware context -- assembled from a living knowledge graph.

How memory retrieval strength works

Click "Recall" to see how spaced repetition builds lasting memory.

100% 50% 0%
0 Days 12mo
Retrievability: 85% Stability: 1.2d Reviews: 2

See your memory space

Engram includes a built-in WebGL graph visualization. Explore connections between memories, search, create, edit, all from the browser.

demo.engram.lol
🧠

Interactive Memory Galaxy

Explore a live demo with sample memories and auto-generated links. Click nodes to inspect. Search, filter by category, create new memories. Keyboard shortcuts for power users.

WebGL Rendering Force-Directed Layout Real-time Search Category Filters
Open Live Demo

Eidolon
Agent activity tracking

Know what your agents are doing. Eidolon captures session starts, task progress, completions, errors, and heartbeats. Query by agent, by project, by time range.

  • Real-time activity feed
  • Session timeline visualization
  • Agent heartbeat monitoring
  • Task state machine tracking
  • Cross-agent coordination events
# Report activity to Eidolon
curl -X POST $EIDOLON_URL/activity \
-H "Authorization: Bearer $EIDOLON_KEY" \
-d '{
"agent": "claude-code",
"action": "task.completed",
"summary": "Deployed v2.1 to prod"
}'

Drop-in hooks.
Automatic memory.

No more "search Engram for..." in every prompt. Install hooks once, memory works automatically. Session context loads at start, memories store on end, relevant context surfaces each turn.

Claude Code Hooks

Four hooks that wire up the entire system:

  • SessionStart -- Load project context
  • UserPrompt -- Search relevant memories
  • PostToolUse -- Report to Mnemonic sidecar
  • Stop -- Store session summary
🛠

Agent-Forge Protocol

Structured reasoning workflow enforced via hooks:

  • spec_task -- Define before coding
  • log_hypothesis -- Record debugging guesses
  • verify -- Check code compiles/runs
  • challenge_code -- Quality self-review
💭

Mnemonic Sidecar

Session-scoped memory companion:

  • /observe -- Capture tool discoveries, auto-flush to Engram
  • /recall -- Hybrid search during session
  • /compress -- Shrink tool outputs, save context
  • /end -- Close session with stats
# Install hooks and forget about manual prompting
cp hooks/simple/*.sh ~/.claude/hooks/
# Or full version with Eidolon, Agent-Forge, credd
cp hooks/full/*.sh ~/.claude/hooks/

Works with
any AI platform.

Connect via MCP protocol, HTTP API, or CLI. Drop Engram into your existing workflow in minutes.

🤖
Claude
🚀
Cursor
🌊
Windsurf
Zed
💬
ChatGPT
💻
Any MCP
MCP Native | HTTP REST API | CLI Tool

Built for
production.

Rust-native performance with zero garbage collection pauses. Local ONNX inference means no API latency.

<15ms
Search Latency
384
Embedding Dims
0
Cloud Dependencies
9
Rust Crates
Memory Recall Accuracy 94%
FSRS Stability After 7 Days 87%
Search Relevance (MRR@10) 0.91
Graph Entity Resolution 89%

Drop in
anywhere.

CLI Bash
# Store a memory
engram-cli store "Auth migrated to JWT" \
  --category decision \
  --importance 9

# Search
engram-cli search "auth setup"

✔ 3 results in 12ms
HTTP API curl
# Store via API
curl -X POST localhost:4200/store \
  -H "Authorization: Bearer eg_..." \
  -d '{"content":"Auth migrated"}'

# Contextual recall
curl -X POST localhost:4200/context \
  -d '{"query":"auth setup"}'
Quick Start 30 seconds
# Clone and build
git clone github.com/Ghost-Frame/Engram-rust
cargo build --release

# Run server
./target/release/engram-server

✔ Ready on :4200

Try it yourself.

CLI playground, live memory graph, and store demo. All running in your browser.

CLI Playground

engram-cli
# Type a command and press Enter
# Try: store, search, list, help
$

Store a Memory

Get started in
30 seconds.

One command. Full memory. Your hardware.

$ cargo build --release

127.0.0.1:4200 · ./data · libsql + ONNX · No external dependencies