zaungast
A read-only, offline MCP server for Microsoft Teams that searches chats, reads conversations, and finds people directly from the local on-disk cache with no cloud API.
README
zaungast
<img src="https://raw.githubusercontent.com/mbe24/zaungast/main/assets/logo.png" alt="zaungast logo: a hatted figure peeking over a fence with binoculars" width="150" align="right">
zaungast (German: someone who watches over the fence without joining in) is a read-only, offline MCP server for Teams — search chats, read conversations, surface trending topics, and find people straight from the local on-disk cache, with no Graph API, no cloud, and no credentials, and token-economical output for coding agents (Claude Code, Claude Desktop, …).
The new Teams client stores your chats in a local on-disk database; zaungast reads a copy of it directly and serves it over MCP, so your agent can pull in Teams context — "what did Anna say about the deploy", "catch me up on the platform channel", "what's my team on about this week" — without you copy-pasting, and without any cloud API.
- Local & offline — reads the on-disk Teams cache. No MS Graph API, no network calls.
- No credentials — nothing to log in to, no tokens, no permissions to grant.
- Read-only & safe — the Teams files are only ever read/copied, never written, locked, or modified. It cannot corrupt your Teams data.
- Token-economical — every tool returns compact, shaped output, never bulk dumps.
- Zero-config — auto-discovers the local Teams database; just register and go.
⚠️ Windows only (new Teams / WebView2), Node.js ≥ 22.5. Teams on macOS uses a different storage engine and isn't supported yet — see requirements. Not affiliated with or endorsed by Microsoft. It reads your own local data on your own machine.
Installation
npx -y zaungast fetches and runs the server, so registering it in your MCP client is the
whole install. No environment variables are needed in the common case — the local Teams
database is auto-discovered.
Claude Code
Normal case — the Teams database is auto-discovered, so no variables are needed:
claude mcp add zaungast -- npx -y zaungast
Add --scope user to make it available in every project. Verify with claude mcp list
(expect zaungast … ✓ Connected), then open a new session.
Only if auto-discovery fails or you have multiple profiles, set TEAMS_LEVELDB_DIR
explicitly (the -e flag goes before the --) — see
finding the Teams database folder:
claude mcp add zaungast \
-e TEAMS_LEVELDB_DIR="C:\Users\me\AppData\Local\Packages\MSTeams_8wekyb3d8bbwe\LocalCache\Microsoft\MSTeams\EBWebView\WV2Profile_tfw\IndexedDB\https_teams.microsoft.com_0.indexeddb.leveldb" \
-- npx -y zaungast
Codex
Add to ~/.codex/config.toml (or run codex mcp add zaungast -- npx -y zaungast):
[mcp_servers.zaungast]
command = "npx"
args = ["-y", "zaungast"]
Claude Desktop / other clients
Add to your client's MCP config (claude_desktop_config.json, .mcp.json, …):
{
"mcpServers": {
"zaungast": { "command": "npx", "args": ["-y", "zaungast"] }
}
}
From-source setup and other clients are covered in the installation docs.
Tools
| Tool | What it does |
|---|---|
list_conversations |
Your Teams sidebar — newest conversations, or filter by kind/participant/title/time. |
read_messages |
One conversation's messages in story order (window / cursor / around a hit). |
search |
Full-text search + filters (from, in, kind, mentions_me, has-attachment, date). |
top_topics |
Distinctive/trending topics over a window, vs your baseline, with an example each. |
find_person |
Resolve a name/nickname to a canonical person + handle, with contact stats. |
describe_schema |
Recovery tool: propose a field mapping when a Teams update changes the DB layout. |
Full reference: tools documentation.
Environment variables
All optional — zaungast works with no configuration. Pass them via your MCP client's env
block (as above).
| Var | Default | Notes |
|---|---|---|
TEAMS_LEVELDB_DIR |
auto-discovered | Path to the …\IndexedDB\https_teams.microsoft.com_0.indexeddb.leveldb directory. Set only if discovery fails or to pin a profile. |
ZAUNGAST_INCREMENTAL |
copy-reuse |
Refresh mode: copy-reuse (faster) or reparse (simpler). |
ZAUNGAST_DB_DIR |
unset | Read a static copy of the database directly (offline analysis); skips discovery and live refresh. |
Privacy & safety
- Stays local — no network calls; reads data already on your machine and serves it to your local agent over stdio.
- Cannot harm Teams — no code path writes to, locks, or memory-maps the Teams directory; only read-and-copy.
- Images/files are URL-only — chat images live in Teams' cloud behind auth; zaungast notes that an attachment exists but never fetches it and never handles credentials.
More in the privacy & safety docs.
Documentation
Full documentation — tools reference, how it works, configuration, privacy, troubleshooting, and development — lives at https://zaungast.readthedocs.io/.
License
Apache 2.0 © Mikael Beyene. Not affiliated with or endorsed by Microsoft. "Microsoft Teams" is a trademark of Microsoft Corporation; this project only reads your own local data.
Recommended Servers
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.
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.
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.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
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.
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.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
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.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.