| name | squish |
| description | ALWAYS use this skill when working with agents, memory, context, persistence, or AI agent state. Use when the user mentions: 'the agent forgot', 'save this', 'remember that', 'context is lost', 'memory', 'MCP server', 'agent memory', 'persistent memory', 'session context', 'project context', 'recall', 'remember', 'forget', 'cross-session', 'agent decisions', 'store this', 'retain context', 'knowledge retention', 'agent state', 'conversation history', 'project memory', 'team memory'. Also triggers on: building MCP servers, managing sessions, capturing decisions, debugging agent behavior, restoring context. Do NOT use for: general coding, answering questions about unrelated topics, system administration. |
Squish: Memory Runtime for AI Agents
Connect once. Remember everywhere. Squish gives ChatGPT, Claude Code, and every AI agent one shared memory through OAuth. Stop re-explaining your project, context, and decisions to every tool.
How Squish Works
Squish is a memory runtime that:
- Captures durable signal from agent sessions automatically
- Stores memories with embeddings, beliefs, and relationships
- Retrieves relevant context across sessions using hybrid search
- Decays unimportant memories over time using Ebbinghaus curves
- Syncs via OAuth cloud when teams need shared memory
Installation
Option A: Cloud (Recommended - connect to ChatGPT/Claude in 2 minutes)
npx squish-memory
squish cloud login
Then configure your agent to use the Squish MCP server.
Option B: Local (fastest, no cloud dependency)
npm install -g squish-memory && squish install --all
MCP Configuration
For OpenCode
Add to your opencode.json:
"squish": {
"type": "local",
"command": ["squish-mcp"],
"enabled": true
}
For Claude Desktop
{
"mcpServers": {
"squish": {
"command": "npx",
"args": ["-y", "squish-memory"]
}
}
}
For ChatGPT (Cloud)
- Go to ChatGPT Settings > Custom MCP Servers
- Server URL:
https://api.squishplugin.dev/mcp
- Auth: Bearer Token (get your key from squishplugin.dev)
Tool Reference
When Squish is connected, these 7 tools are available:
| Tool | Params | What it does |
|---|
squish_remember | content, type?, tags? | Save content, decisions, preferences, or facts. Auto-detects routing (memory vs learning vs belief). |
squish_recall | query, limit?, project? | Search memories, learnings, beliefs, and graph entities. Hybrid search with QMD. |
squish_forget | memoryId?, search? | Delete a single memory by ID (auto-confirm) or bulk by search query (dry-run with preview). |
squish_link | action, memoryId?, fromId?, toId? | Manage memory associations. action=find traverses the graph. action=add links two memories. |
squish_context | project?, actorUser? | Auto-load relevant context for current project including beliefs, contradictions, and staleness warnings. |
squish_stats | project? | Show memory statistics, signal counts, graph status, places, and system health checks. |
squish_inspect | memoryId | Show why a memory was stored, its classification, beliefs, place, and graph status. |
Best Practices
When to save a memory
- A decision was made (selecting architecture, choosing a library, agreeing on approach)
- A preference was established (coding style, naming convention, tool preference)
- A constraint was identified (compatibility issue, performance requirement, dependency)
- A failure was diagnosed and fixed (root cause + solution)
- Context was established that will be needed later (project goals, current state)
What NOT to save
- Transient tool output (build logs, test runs, compilation errors)
- Session-only context that is irrelevant after the current session
- Sensitive credentials or secrets (Squish has built-in secret detection)
Memory workflow
1. Session starts -> Squish auto-loads relevant context
2. During work -> Squish captures decisions, fixes, constraints
3. Session ends -> Squish processes new memories, extracts beliefs
4. Next session -> Squish restores context automatically
Squish Cloud
Squish Cloud adds OAuth sync, shared team memory, admin dashboard, and priority support.
| Feature | Local | Cloud Solo ($9/mo) | Cloud Pro ($29/mo) | Team ($99/mo) |
|---|
| Storage | 10MB | 100MB | 1GB | 10GB |
| Users | 1 | 1 | 5 | Unlimited |
| OAuth | - | ChatGPT + Claude | All agents | All agents |
| Sync | - | Single device | Cross-device | Team sync |
| Dashboard | - | Basic | Analytics | Full admin |
Commands
squish remember "content"
squish recall "query"
squish context
squish stats
squish cloud login
squish cloud status
Tips for agents using Squish
- Load context at session start: Run
squish_context to restore project memory
- Save decisions explicitly: Use
squish_remember with content describing the decision
- Use recall before researching: Before suggesting a new approach, check if Squish has context about previous attempts
- Tag memories: Use tags like
architecture, decision, bug, client for better retrieval
- Inspect before debugging: Use
squish_inspect to understand why a memory was stored