nelly-elephant-mcp

nelly-elephant-mcp

Search and resume past Claude Code conversations via MCP. Reads local JSONL session files, runs locally, MIT licensed. An elephant never forgets.

Category
Visit Server

README

nelly-elephant-mcp

An elephant never forgets — and neither should your AI.

Search and resume past Claude Code conversations via the Model Context Protocol (MCP).

Read the full story →

Install

npm install -g nelly-elephant-mcp

Or install from source:

git clone https://github.com/SpicyRiceCakes/nelly-elephant-mcp.git
cd nelly-elephant-mcp
npm install && npm run build
npm install -g .

Configure

Where Nelly looks

Claude Code stores all session transcripts — from every project — in one place:

~/.claude/projects/
  ├── -Users-you-project-alpha/     ← sessions from project alpha
  │   ├── abc123.jsonl
  │   └── def456.jsonl
  ├── -Users-you-project-beta/      ← sessions from project beta
  │   └── ghi789.jsonl
  └── ...                           ← every project you've ever opened

Nelly searches all of them at once. It doesn't matter which project you're in when you ask — she searches everything.

Recommended: Global config

Because Nelly searches across all projects, install her globally so she's available in every session. Add to ~/.claude/mcp_settings.json:

{
  "mcpServers": {
    "nelly": {
      "command": "nelly-elephant-mcp",
      "args": []
    }
  }
}

Alternative: Per-project config

If you only want Nelly in specific projects, add to your project's .mcp.json instead:

{
  "mcpServers": {
    "nelly": {
      "command": "nelly-elephant-mcp",
      "args": []
    }
  }
}

Even in per-project mode, Nelly still searches all sessions. The only difference is whether she's available to call.

Restart Claude Code after configuring. Nelly is ready.

Trunk Modes

Nelly has two modes — just say which one you want:

  • 🐘 Long Trunk — the full Nelly experience. She'll talk you through the search, ask follow-up questions, and celebrate when she finds it. An elephant never forgets.
  • 🐘 Short Trunk — just the results. No personality, no narration. Date, project, snippet, resume command. Done.

Say "Nelly long trunk" or "Nelly short trunk" to switch anytime.

Tools

Tool What it does
nelly_instructions Returns usage guide and system status.
nelly_search Search past sessions by keyword. Supports date range and project filters.
nelly_context Get expanded context from a specific session.
nelly_recent List most recent sessions.

How it works

Claude Code stores session transcripts as JSONL files in ~/.claude/projects/. Nelly searches these files, extracts human-readable text, and returns matching sessions with claude --resume commands.

You can filter by project name or date range if you want to narrow the search. But by default, Nelly searches everything — that's the point.

The AI uses a "spiral search" pattern — starting broad, trying variations, and asking for hints rather than giving up after one attempt.

Privacy

Elephants have trunks, not phones.

What Nelly does: Reads local files on your machine. Zero network calls. Zero telemetry. Zero data collection. Nelly doesn't send data anywhere. Read the source — it's four files.

What you should know: Nelly runs inside Claude Code. When she finds a matching session, the snippets and project names she returns become part of your current conversation for Claude to process — the same way anything you discuss in a Claude session does. Nelly isn't sending data anywhere extra — it's just that whatever she finds becomes part of the chat, like reading a file or pasting text.

In plain English: Nelly searches locally, but the results become part of your conversation. For sensitive searches, use grep directly in your terminal instead — that stays fully on your machine.

Use Nelly for moments of need, not as a browsing tool. Every search puts past session content into your current conversation. Search when you need to find something specific, not to casually explore your history.

Never use Nelly to search for passwords, API keys, or secrets. If a past session contains a leaked credential, searching for it with Nelly copies it into a second session — doubling the exposure. Use grep directly in your terminal instead:

grep -r "your-search-term" ~/.claude/projects/

That stays fully local. No API, no cloud, no second copy.

Compatibility

Version Platform Status
v1.0 Claude Code (macOS, Windows, Linux) Live
v1.a Codex CLI PRs welcome
v1.b Gemini CLI PRs welcome
v1.c Cursor Research needed
v1.d Windsurf Research needed
v1.e Cline Research needed

Contributing

We built the Claude Code adapter. Want to make one for your tool?

  • Fork the repo and look at src/search.ts — that's the pattern
  • Adapters need to find session files and parse them for your tool's format
  • PRs welcome

Origin

Born from a real need — couldn't find a fix from a previous session. Knew it was somewhere in the conversation history but couldn't remember which one. Searched with grep, found it, realized this should be a tool. Built it that night.

Read the full story →

License

MIT


Built by SpicyRiceCakes x Claude

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured