Memory infrastructure for AI workflows

AI that remembers.
Context that compounds.

Unbidden builds memory infrastructure for AI development workflows. Your AI gets smarter the longer it works with you.

Try Engram → View on GitHub

# Connect to your AI editor in one step

engram mcp-serve

 

# Every session starts with full context.

# Week-1 decisions retrieve in week 12.

Every AI session starts from zero. Unbidden fixes that.

Product

Built for developers who work in long-running AI sessions

One product, focused on the problem that kills most AI-assisted work: the context window.

Engram — Persistent AI Memory

A context engineering memory system for AI agents. Retrieves all relevant context and only relevant context — reducing token consumption while keeping your AI grounded in what you've already decided.

MCP server REST API CLI Local or hosted OpenAI-compatible Claude Code Cursor Windsurf
Learn more →

How it works

Three steps. No ceremony.

Engram runs in the background. You get the benefit without changing how you work.

1

Connect to your AI editor via MCP or REST API

One config block in Claude Code, Cursor, Windsurf, or any MCP-compatible client. Done.

2

Engram extracts what matters from every session

Decisions, constraints, implementation details — extracted at session end and merged into a persistent knowledge graph.

3

Next session, retrieve exactly what's relevant

Not everything, not nothing. Semantically scored, temporally weighted. Your AI picks up exactly where it left off.