♊
Gemini CLI+Stompy

Google's terminal AI with a memory it can finally keep

Because every new Gemini session shouldn't feel like a first date

The Problem

Gemini CLI is fast. It's smart. It understands your code in ways that make you feel slightly inadequate.

And then you close your terminal.

The next session, Gemini greets you like you've never met. Your project structure? Unknown. Your conventions? A mystery. That three-hour debugging session where you finally tracked down the race condition? Ancient history that nobody recorded.

Google gave Gemini a 1M token context window. Impressive. But context windows don't survive session restarts. Your AI has the attention span of a goldfish with a very, very large short-term memory.

How Stompy Helps

Stompy gives Gemini CLI persistent memory that survives every session restart — via three integration paths that all hit the same memory store.

MCP: Native tool integration via Gemini's settings.json. Gemini calls lock_context and recall_context as naturally as it reads your files.

CLI: Shell commands — stompy lock, stompy recall, stompy search. Works from any terminal, any script, any CI/CD pipeline.

REST API: HTTP endpoints at /api/v1/agent/memory/*. For custom scripts, webhooks, or anything that speaks HTTP.

Three paths, one brain. Context locked via MCP is recallable via CLI or API and vice versa. Switch tools, switch sessions, switch machines — your project memory follows.

Integration Walkthrough

1

Path 1: Connect via MCP (native integration)

Add Stompy to Gemini CLI's MCP configuration for automatic tool access.

// ~/.gemini/settings.json
{
  "mcpServers": {
    "stompy": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.stompy.ai/sse"],
      "env": {
        "AUTHORIZATION": "Bearer YOUR_STOMPY_TOKEN"
      }
    }
  }
}
2

Gemini uses Stompy tools automatically

Once configured, Gemini calls lock_context and recall_context as native MCP tools.

# Session 1: Gemini learns your architecture
gemini "explain our authentication flow"
# Gemini saves: lock_context(topic="auth_flow", content="JWT with RS256...")


# Session 47: Gemini remembers
gemini "add rate limiting to the auth endpoint"
# Gemini recalls auth_flow automatically — no re-explaining needed
3

Path 2: Use Stompy CLI directly

Shell commands that work from any terminal or script. Same memory store as MCP.

# Lock a memory from your terminal
stompy lock --topic "api_conventions" \
  --content "REST endpoints use /api/v1/ prefix, snake_case, 201 for creates"


# Recall in any session
stompy recall --topic "api_conventions"


# Semantic search across all memories
stompy search "how do we handle authentication"
4

Path 3: Use the REST API

HTTP endpoints for custom integrations, CI/CD, or any tool that speaks HTTP.

# Store a memory via REST API
curl -X POST https://api.stompy.ai/api/v1/agent/memory/store \
  -H "Authorization: Bearer $STOMPY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"topic": "deploy_config", "content": "Production uses 3 replicas..."}'


# Recall a memory
curl -X POST https://api.stompy.ai/api/v1/agent/memory/recall \
  -H "Authorization: Bearer $STOMPY_TOKEN" \
  -d '{"query": "deployment configuration"}'
5

Cross-agent workflow: Gemini + Claude

Memories saved by one AI are instantly available to another. Same project, same brain.

# Morning: Gemini designs the API schema
gemini "design the user settings API"
# → lock_context(topic="settings_api", content="GET/PUT /settings...")


# Afternoon: Claude implements it — with full context
claude "implement the settings API we designed this morning"
# → recall_context(topic="settings_api") — gets Gemini's design

What You Get

  • Three integration paths: MCP, CLI, and REST API — all hitting the same memory store
  • Cross-agent memory: context saved by Gemini is recalled by Claude, Codex, or any tool
  • Semantic search finds relevant context even when you don't remember the exact topic
  • Session handovers survive terminal restarts, machine switches, and tool changes
  • Project isolation keeps separate codebases completely independent

Ready to give Gemini CLI a memory?

Join the waitlist and be the first to know when Stompy is ready. Your Gemini CLI projects will never forget again.

Related Integrations

Codex CLI

AI Coding Assistants

OpenAI's terminal agent with cross-session memory

Learn more

Claude Code

AI Coding Assistants

Your terminal companion with perfect recall

Learn more

Cursor

AI Coding Assistants

The AI-first IDE that never forgets

Learn more
View all integrations