The memory database for AI agents.

Graph memory, hybrid search, and an optional tamper-evident journal.
Runs on SQLite. No server required. No LLM in the core write path.

Quick start

Two calls: remember to store, recall to search. Auto-embed. No IDs, no vector plumbing.

import asyncio

from engrava import SqliteEngravaCore


async def main():
    # from_config opens and owns the connection (schema + PRAGMAs applied).
    async with await SqliteEngravaCore.from_config("engrava.yaml") as store:
        # Store a memory in one call — no IDs, no record to assemble.
        await store.remember("User prefers concise answers")
        await store.remember("User works in Berlin")

        # Hybrid search: FTS5 + vector + recency + priority + graph.
        result = await store.recall("what does the user prefer?")
        for thought_id, score in result.results:
            print(score, thought_id)


asyncio.run(main())

Full API reference in the docs.

What you get

Twelve primitives. One pip install. No required service, no per-operation meter, no egress unless you configure a remote provider.

• Graph Memory

  7 typed edge kinds in the same SQLite store.

• Embedded

  One pip install. Runs inside your Python process. No server.

• Hybrid Search

  FTS5 + vector + recency + priority + graph. 5-signal fusion.

• Lifecycle Management

  CREATED → ACTIVE → DONE → ARCHIVED.

• Dreaming

  Algorithmic memory consolidation, built in.

• Audit Trail

  Optional SHA-256 hash-linked journal. Tamper-evident.

• Batteries Included

  Graph, search, consolidation, and optional audit journaling — one local package.

• MindQL

  FIND, COUNT, SELECT. No Cypher.

• 5 Embedding Providers

  Local, OpenAI, Ollama, HuggingFace. Auto-embed.

• Bi-temporal

  Valid-time on every fact. Time-travel queries with valid_at, valid_now. Invalidate without deleting.

• MCP Server

  Connect your agent — Claude Desktop, Claude Code, Cursor. Standalone server: uvx engrava-mcp. Read-only mode available.

• Multi-service

  Isolated memory per service — one database file each, managed together.

How engrava compares

Graph-first. Local by default. No required service. Everything in one developer workflow.

How dreaming works

Install & configure

One pip install. Optional extras for embedding backends. One YAML file for the rest.

# Basic
pip install engrava

# With local embeddings (sentence-transformer)
pip install engrava[embeddings-local]

# With OpenAI-compatible embeddings (OpenAI, Azure, Groq, vLLM, LiteLLM)
pip install engrava[embeddings-openai]

# Alt: Ollama (local LLM server) or HuggingFace Inference API
pip install engrava[embeddings-ollama]
pip install engrava[embeddings-hf]

# engrava.yaml
database:
  path: "./agent.db"

embeddings:
  provider: "sentence-transformer"    # or "openai-compatible", "ollama", "huggingface"
  model: "all-MiniLM-L12-v2"
  auto_embed: true

extensions:
  dreaming:
    enabled: true
    promote_threshold: 0.75
    signals:                # name -> weight (sum ~1.0)
      recency: 0.25
      staleness: 0.20
      confirmation: 0.20
      confidence: 0.15
      frequency: 0.20
    gates:
      min_confirmations: 2
      max_promoted_per_run: 20

journal:
  enabled: true   # optional tamper-evident journal (SHA-256 hash chain)

Full configuration reference in the docs.

Built from research.