squatch-harbor-mcp
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@squatch-harbor-mcp search harbor for when I discussed authentication"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
squatch-harbor-mcp
Search your own past Claude sessions, locally. Indexes ~/.claude/projects/*.jsonl into a
local SQLite FTS5 file and searches it — "when did I set up auth?", "how did we solve X?".
No cloud, nothing leaves your machine. Stdlib only (no numpy). An MCP
server, so it works in Claude Desktop (Chat) and Claude Code.
Run it (zero install — uvx fetches + runs)
uvx --from git+https://github.com/squatch-c-c/squatch-harbor-mcp squatch-harbor-mcpRequires uv (curl -LsSf https://astral.sh/uv/install.sh | sh).
Related MCP server: MCP Claude Context Server
Add to Claude Desktop
~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"harbor": { "command": "uvx", "args": ["--from", "git+https://github.com/squatch-c-c/squatch-harbor-mcp", "squatch-harbor-mcp"] }
}
}Restart Claude, then: "ingest my harbor, then search it for when I set up the egress guard."
Tools
Tool | What it does |
| Index your sessions (incremental; run first). |
| Top matches — session, date, role, snippet. |
| Index size + date range. |
Local-only. Index lives at ~/.harbor/harbor-mcp.db. Part of the Squatch system.
Available Tools
3 toolsharbor_ingestA
Index your local Claude session transcripts so you can search them. Incremental by default (only re-reads changed files); pass rebuild=true to start fresh. Defaults to ~/.claude/projects.
Inputs: sessions_dir, rebuild.
harbor_searchA
Search your indexed Claude history for past work ('when did I…', 'how did we solve…'). Returns the top matches with session, date, role, and a snippet. Run harbor_ingest first.
Inputs: query (required), k. Returns: result
harbor_statsA
How big the harbor index is (message count, file count, date range) — or that it's not built yet.
Inputs: no named parameters.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 3 tools
Each tool serves a unique purpose: ingest indexes files, search queries the index, and stats reports index metadata. No functional overlap.
All tools follow a consistent 'harbor_verb' pattern (harbor_ingest, harbor_search, harbor_stats), using snake_case and clear verbs.
With 3 tools, the set is minimal but perfectly scoped for the domain: index creation, querying, and status checking. No unnecessary tools.
The tool surface covers the full lifecycle: ingest (create/update), search (read), and stats (monitor). A tool to delete or clear the index is not provided but can be achieved via rebuild flag, so minor gap.
Related MCP Connectors
Search your AI chat history (ChatGPT, Claude, Codex) from any MCP client. Remote, private, read-only
Shared memory MCP server for Claude Code, Codex, Gemini CLI and Cursor. Cited, editable.
Hosted MCP memory: save sessions/decisions once, search from Claude, Cursor, ChatGPT. EU-hosted FTS.
Related MCP Servers
- AlicenseAqualityBmaintenanceMCP server for browsing, searching, exporting, and backing up your Cursor AI chat history directly into Claude via natural language.881 npm33MIT
- AlicenseNot gradedqualityDmaintenanceA powerful MCP server that extracts, stores, searches, and analyzes Claude.ai conversations with SQLite database, semantic search, and multiple export formats including Obsidian and PDF.101 PyPI1MIT
- AlicenseAqualityDmaintenanceA zero-dependency MCP server that enables searching and reading local Claude Code and Codex chat sessions, supporting full-text search, grep, and knowledge indexing from chat history.1516 npmMIT
- AlicenseNot gradedqualityDmaintenanceA local MCP server that indexes and searches your Claude Code conversation history with both keyword and semantic search, fully private and running locally.MIT