Skip to main content
Glama
deenesjakoruzh
Ipramking

Engram MCP

by Ipramking
OverviewSchemaSourceRelated ServersDiscussionsAdmin
TypeScript
Remote

🧠 Engram

Portable, verifiable memory for AI agents — on Walrus. Tell your agent a decision, config, gotcha or fact about a teammate; it stores it on Walrus and recalls it across sessions, devices, models — and, in team scope, across your whole team. It stops re-asking for context you already gave it.

Built for the Walrus Memory hackathon (SuiHub Lagos, Aug 2026).

Two surfaces, one idea

  • Engram MCP — add it to Claude Code / Cursor / Codex and your agent gets engram_remember / engram_recall / engram_tidy with personal vs team scope and typed capture rules. This is the reproducible artifact: real Walrus writes, verifiable blob ids.

  • Engram web app — a chat agent that auto-captures durable facts, shows each one's Walrus blob id + inspect link, and has a 🔒 Personal ↔ 👥 Team toggle and a model switcher. Runs with zero keys via an offline mock.

Related MCP server: Archetypal AI MCP Server

The prompt

PROMPT.md is the copy-pasteable system prompt that drives it — the rules for what to remember and when to recall. Paste it into any agent wired to these tools.

Quick start

pnpm install

# The MCP (real Walrus memory in your agent):
pnpm login:memwal        # one-time browser sign-in
# → add mcp/engram-mcp.mjs to your client (see mcp/README.md), paste PROMPT.md

# The web app demo (no keys needed):
pnpm dev                 # http://localhost:5173

Proof it works

Memories written live to Walrus mainnet via engram_remember (aggregator returns HTTP 200):

See SUBMISSION.md for the full write-up and VIDEO.md for the demo script.

Layout

PROMPT.md          the copy-pasteable memory protocol (the star deliverable)
mcp/engram-mcp.mjs the MCP server (scope + typed rules) + mcp/README.md
src/, api/         the web app (Vite/React) + its serverless API
scripts/           smoke tests and proof collectors

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Maintenance

ActivityMaintained
ResponsivenessNo issues

Resources

Related MCP Connectors

Related MCP Servers

  • PLUR by plur-ai

    PLURofficial

    plur-ai
    A
    license
    B
    quality
    A
    maintenance
    Open, local-first engram memory for AI agents: read, correct, and delete what your agent remembers. Works across Claude Code, Hermes, OpenClaw, and Cursor over MCP.
    39
    2,812,062 npm
    295
    Apache 2.0

  • Archetypal AI MCP Server

    archetypal-ai
    F
    license
    Not graded
    quality
    C
    maintenance
    Provides persistent memory tools (recall, remember, checkpoint) for AI agents, enabling them to save and restore state across sessions.
    4
    -

  • strata

    rx6ru
    A
    license
    Not graded
    quality
    A
    maintenance
    Enables agents to read and write a durable, append-only memory with verified provenance, supporting search, record capture, and context preparation across sessions and machines.
    MIT