Skip to main content
Glama

Engrams

Persistent Project Memory for AI Assistants

License Python 3.8+

Engrams is an MCP server that gives AI assistants structured, queryable memory for your projects — decisions, patterns, progress, and team rules — so you stop re-explaining your stack in every prompt.

Documentation · Issues


Related MCP server: LumenCore

Installation

Prerequisites

Add to Your MCP Client

Add the following to your MCP settings file (mcp.json or your IDE's MCP configuration):

{
  "mcpServers": {
    "engrams": {
      "command": "uvx",
      "args": [
        "--reinstall",
        "--from",
        "engrams-mcp",
        "engrams-mcp",
        "--mode",
        "stdio",
        "--log-level",
        "INFO"
      ]
    }
  }
}

Works with Roo Code, Cline, Cursor, Windsurf, Claude Code, and any MCP-compatible client.

Workspace detection is automatic — no --workspace_id flag required. Engrams locates your project root per-call using .git, package.json, and similar indicators.

Add Custom Instructions to Your Project

After installing, run engrams init from your project root to scaffold the strategy file for your AI tool:

engrams init --tool roo          # → .roo/rules/engrams_strategy + .roomodes + system prompts
engrams init --tool cline        # → .clinerules
engrams init --tool cursor       # → .cursorrules
engrams init --tool windsurf     # → .windsurfrules
engrams init --tool claude-code  # → CLAUDE.md
engrams init --tool generic      # → engrams_strategy.md

engrams init --list              # See all supported tools

Reducing Token Costs

Every time you start a new AI session, you're re-paying to re-establish project context: your stack, your conventions, your current task. With a large project and frequent sessions, this adds up fast.

Engrams stores that context once in a local SQLite database. Your AI assistant retrieves only what's relevant to the current task — not everything, not nothing.

Before Engrams — you prompt with background context every session:

You: We use FastAPI with SQLAlchemy and PostgreSQL. JWT auth is required on all
     routes. Follow the existing error handling pattern. Now add a /medications
     endpoint that...

After Engrams — the context is already there:

You: Add a /medications endpoint that...

AI: Retrieving relevant context...
    Decision #23: PostgreSQL primary database
    Decision #24: SQLAlchemy for all DB operations
    Decision #7: JWT auth required on all routes
    Pattern #5: Structured error handling

    Implementing with SQLAlchemy, JWT middleware, and structured error responses...

The more you log, the less you repeat. Features like codebase bindings and context budgeting ensure that only the relevant context for each task is loaded — keeping prompt sizes (and costs) low even as the knowledge base grows.


Establishing Team Rules

Engrams has a two-layer governance system: team-level rules that apply to everyone, and individual-level rules for personal workflow preferences.

Log a team decision

You: "The team has decided to use PostgreSQL for the primary database because
     we need ACID guarantees and complex query support. This is a team-level
     architectural decision."

AI: Decision #23 logged:
    Use PostgreSQL for primary database
    Rationale: ACID guarantees, complex query support, mature ecosystem
    Scope: Team Architecture
    Tags: database, architecture

Block conflicting proposals automatically

Once a team rule is in place, Engrams enforces it:

You: "I want to use MongoDB for the analytics module"

AI: ⚠ Governance Conflict Detected

    Your proposal conflicts with Team Decision #23:
    "Standardize on PostgreSQL for all data storage"
    Scope: Core Architecture (Team-level, hard block)

    Would you like to propose an amendment instead?

Allow individual overrides

Individual developers can still log personal workflow strategies without affecting team rules:

You: "When I'm ready to create a PR, compare my branch with main, generate
     a summary of my changes, run autoflake on the API, and fix any critical
     lint errors."

AI: Individual strategy #4 created and saved.

See the Governance documentation for scopes, amendment workflows, and compliance checks.


Learn More

The Engrams documentation site covers all features in detail:


License

Apache 2.0 · Forked from GreatScottyMac/context-portal v0.3.13

A
license - permissive license
-
quality - not tested
B
maintenance

Maintenance

Maintainers
Response time
1wRelease cycle
3Releases (12mo)

Resources

Unclaimed servers have limited discoverability.

Looking for Admin?

If you are the server author, to access and configure the admin panel.

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/stevebrownlee/engrams'

If you have feedback or need assistance with the MCP directory API, please join our Discord server