OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Agent Memory & RAG

Engram — MnemoDev

Engram is a self-hosted retrieval-augmented memory service for AI agents. It stores observations, indexes them in ChromaDB, retrieves context by semantic similarity and recency, merges near-duplicates, and expires or prunes entries using time-to-live rules.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Engram — MnemoDev — OpenViST project profile in Agent Memory & RAG.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Engram when an agent needs to carry observations across cycles and retrieve them by semantic relevance, with memory expiry and cleanup handled by the service.

Who it is for

  • Developers adding persistent observation storage and recall to AI agents.
  • Teams that need categorized agent memories with automatic expiry and duplicate merging.

Documented uses

  • Store observations from an agent's work between decision cycles.
  • Search for relevant prior context before an agent makes a decision.
  • Inspect memory-store statistics and prune expired entries.
Before you choose

Requirements & limits

The project is presented as a public GitHub repository under the MIT license, with self-hosted setup instructions.

  • The documented primary stack is Bun 1.2 and ChromaDB; an in-memory backend is also described.
  • Search ranking blends cosine similarity at 85% with recency at 15%, and the weights can be changed in the repository.
  • Default TTLs vary by category: pattern 90 days, warning 60, outcome 180 and context 30.
  • Search filters expired entries; the separate prune operation scans the full collection and is O(n).
  • The README describes a console for memory intake, recall results, store health and event logs.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Start the local development server

Install Engram's dependencies and start its development server with ChromaDB.

Before you start

  • Bun 1.2
  • Docker Compose for the ChromaDB setup

Steps

  1. Clone the Engram repository and change into its directory.

  2. Run bun install to install dependencies.

  3. Start ChromaDB with docker-compose up chromadb -d.

  4. Run bun run dev.

Expected outcome

The README's development-server command is running with the ChromaDB service started.

Guide limitations

  • The README also describes an in-memory backend selected with STORE_BACKEND=memory in .env, which does not require Docker.
  • These are repository setup instructions, not a verified execution result.
Documentation 1 ↗
Source-based practical guide

Use the documented memory API

Submit an observation, search for relevant memories, prune expired entries and request store statistics.

Before you start

  • An Engram server running with a configured memory backend
  • An API client capable of sending the documented HTTP methods and JSON bodies

Steps

  1. Send POST /memories with a category and content; the example also includes a pool address and tags.

  2. Send POST /search with a query and optional topK and category fields.

  3. Send POST /prune to remove expired entries.

  4. Send GET /stats to request store statistics.

Expected outcome

The documented API provides routes for storing and searching observations, pruning expired entries and requesting statistics.

Guide limitations

  • The README gives route examples but does not specify a base URL, authentication method or deployment procedure.
  • Pruning scans the collection and is O(n) in collection size.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
14 October 2026
Original website title
MnemoDev/Engram · Official homepage title · retained in its source language
How we review sourcesSuggest a correction