MCP 服务器

Squish Memory

Connect once. Remember everywhere. Squish gives ChatGPT, Claude Code, and every AI agent one shared memory. Stop re-explaining your project to every tool.

README

Squish -- Connect once. Remember everywhere.

Squish gives ChatGPT, Claude Code, and every AI agent one shared memory.

Without Squish:

Agent restarts. Context disappears.
You explain everything again.

With Squish:

Restart the agent.
Continue working.

Install (30 seconds)

npm install -g squish-memory && squish install --all

Zero config. Zero API keys.

# Show current project context
squish context --json

# Save something explicitly
squish remember "We chose PostgreSQL for team mode" --type decision

# See what the agent knows about a topic
squish recall "project decisions"

The problem: agents forget everything

Every time you start a new session, your AI agent starts from zero. It does not remember the architecture decisions you made last week, the config you spent an hour debugging, or the preference you mentioned yesterday.

You re-explain. Every. Single. Time.

Squish is the memory that persists between sessions. Across agents. Across machines.

How it works

Connect -- OAuth login once in any agent (ChatGPT, Claude Code, OpenCode, Cursor, etc.)
Remember -- Squish auto-captures decisions, constraints, and context as you work
Recall -- When you restart any connected agent, it picks up right where you left off

One memory runtime. Works everywhere.

Why Squish

Most memory tools need a second LLM for embeddings and retrieval. That means extra API costs ($10-100+/mo per agent), latency (500ms+ per call), and more infrastructure to manage.

Squish uses local embeddings by default. No forced LLM dependency.

1-5ms embedding latency (not 500ms+)
$0 runtime cost in local mode
Optional LLM for enhanced extraction and Cloud features

Feature	Most Memory Tools	Squish
Embedding latency	200-1000ms	1-5ms
Local mode cost	$10-100+/mo	$0
LLM required	Yes	Optional
MCP server	Sometimes	Built-in (15 tools)
Cloud sync	Enterprise-only	$9/mo

Features

Squish works with any MCP-compatible agent -- Claude Code, Cursor, OpenCode, Cline, VS Code, Windsurf, Goose, Gemini CLI, Aider, ChatGPT, and more.

Memory intelligence:

Auto-captures decisions, constraints, and preferences as you work
Restores relevant context when an agent restarts
Handles contradictions and temporal facts with expiration
Graph-boosted retrieval across sessions

Interfaces:

CLI: squish remember, recall, inspect, context, stats
MCP Server: 15 tools for any MCP client
Web UI: Local dashboard at localhost:37777
Cloud Dashboard: Analytics and management at squishplugin.dev

Storage:

SQLite (local, default) or PostgreSQL (team mode)
Hybrid retrieval: keyword + semantic similarity
AES-256-GCM encryption for sensitive memories
Places routing: organize memories by context

Squish Cloud

Persistent memory across ChatGPT, Claude Desktop, Claude Code, and local agents. One account, synchronized everywhere.

  ChatGPT          Claude Desktop     Claude Code       Local Agents
 [OAuth 2.1]       [OAuth 2.1]     [Streamable HTTP]  [MCP / CLI]
      +-------------------+---------------+------------------+
                          |
                  Squish Cloud API
                          |
                 [PostgreSQL + Encrypted Storage]
                          |
                  Admin Dashboard & Analytics

Cloud features: OAuth 2.1 + PKCE login, cross-platform sync, team workspaces, admin dashboard, priority support.

Pricing

Tier	Price	Storage	Users
Local	Free	Local SQLite	1
Cloud Solo	$9/mo	50 MB synced	1
Cloud Pro	$29/mo	250 MB synced	1
Team	$99/mo	1 GB shared	Up to 10

Quick Start (Cloud)

npm install -g squish-memory
squish cloud login        # Opens browser for OAuth -- done

Then add to any MCP client:

{
  "mcpServers": {
    "squish-cloud": {
      "type": "url",
      "url": "https://api.squishplugin.dev/mcp",
      "headers": {
        "Authorization": "Bearer <your-token>"
      }
    }
  }
}

Architecture (brief)

Agent Action -> [Filter & Store] -> SQLite/PostgreSQL
                  +                         +
            [Memory Pipeline]        [Hybrid Retrieval]
           auto-capture, dedup,    keyword + semantic search
           graph relationships     with RRF fusion

Two-tier memory pipeline:

Capture filters noisy tool output and promotes what matters -- decisions, constraints, preferences
Retrieval combines keyword and semantic search to find the right context when you need it
Graph connects related memories across sessions for smarter recall
Places route memories into buckets so retrieval stays focused