Best for
- Users who use multiple AI clients and want unified data access
- Those wanting a local-first personal infrastructure
- People who want to easily wire up RPA and automation via MCP
A
autoMate
Community
A smart NAS for AI: notes, files, reminders, memory, 40+ tools, plug into any AI client.
GitHub source repository ↗
34FMRS · D
autoMate is a locally-run personal data and tool layer that provides tools to various AI clients (OpenClaw, Claude Desktop, Cursor, Cline, etc.) via MCP-over-HTTP. It includes a notes system, a file vault (with configurable storage path), hybrid retrieval using BM25, reminders with Web Push, cross-session memory, audio transcription, local executors (shell, scripts, desktop control, browser automation), and 31 SaaS integrations and 25 LLM providers. It can be used as a tool source for another AI client or standalone via its built-in web chat.
Tools
search.find
Hybrid retrieval over notes and files (SQLite FTS5 BM25).
notes.read
Read a markdown note.
notes.write
Create or update a markdown note.
files.list
List files in the vault.
files.write
Write a file to the vault.
reminders.create
Create a reminder and push it to your phone via Web Push.
memory.get
Read a key-value fact from long-term memory.
memory.set
Write a key-value fact to long-term memory.
audio.transcribe
Transcribe voice to text (Tencent ASR or OpenAI Whisper).
shell.exec
Execute a shell command on the host.
script.run
Run a Python, Bash, or Node script.
desktop.control
Control the desktop via pyautogui.
browser.navigate
Control a browser via Playwright.
bx.execute
Control your real browser via the Chrome extension.
automate
Run autoMate's own agent loop.
Setup
After installation, open Settings → Connect to AI clients and click 'Copy install text'. Then follow the instructions for your chosen client (OpenClaw, Claude Desktop, Cursor, Cline, or generic MCP) to paste it into the client's config.
Fit and risk
Not for
- Users who don't want a local server
- Security-sensitive users who are uncomfortable exposing shell.exec
- Those expecting a hosted/managed solution
Required permissions
- Network access (defaults to 127.0.0.1)
- File system read/write (configured storage path)
- System command execution (shell.exec and script.run)
- Browser automation (Playwright and Chrome extension)
- Audio input for transcription
Risks and side effects
- The MCP endpoint is protected by a Bearer token, but leaking the token could allow arbitrary code execution via shell.exec.
- The local server could be exploited by malicious prompts to execute dangerous commands.
- API keys and tokens are encrypted with Fernet, but the key file must be strictly protected.
- autoMate Cloud is opt-in and sends no data unless AUTOMATE_CLOUD_URL is set.
- Using Web Push could expose personal devices.
- SaaS integrations require OAuth tokens, which are stored encrypted.
- Ensure you trust the AI clients you configure with MCP, as they can call shell.exec.
- Audit logs and recommended firewall rules help mitigate risks.
Troubleshooting
- Check that the server is running at http://127.0.0.1:8765
- Verify the token in your MCP configuration matches the one generated in Settings → Channels
- Make sure your client has been restarted to pick up the new MCP server
- Ensure the storage path exists and is writable
- Look at server logs for error messages
Use cases
Use as a personal data warehouse for any AI client (notes, files, memory).
Persist memory across sessions so AI remembers your preferences.
Automate local tasks via natural language (shell, scripts, desktop, browser).
Receive reminders and push notifications.
Manage tool calls across multiple SaaS platforms.
Supported clients
Claude DesktopFull support
OpenClawFull support
CursorFull support
ClineFull support