Channels

v4.5.7: the architecture is now: autoMate is a tool source. OpenClaw (or Claude Desktop, Cursor, Cline, ...) stays the agent; autoMate plugs in via MCP and gives that agent access to the user's notes, files, reminders, memory, audio transcription, and 30+ other tools. The agent decides when to call autoMate.

This replaces the v4.5.6 "autoMate intercepts all messages" design, which conflated channel handling with reasoning. The v4.5.6 inbox endpoint is still shipped for non-MCP gateways (n8n, custom scripts) but is no longer the recommended path.

Two ways to use autoMate

1. As a tool inside another AI client (the main path)

WeChat / Telegram / etc.       OpenClaw / Claude / Cursor
        ▼                              ▼
        └─── (channel plugin) ────►   agent ────► MCP ────► autoMate
                                                  │            │
                                                  ▼            ▼
                                           tools live      files / notes /
                                           there too       memory /
                                                           reminders / agent

The user talks to OpenClaw (or whatever client) on their preferred chat platform. The agent in OpenClaw decides "this needs the user's files / notes / schedule" and calls autoMate's tools. autoMate returns the data; the agent composes the reply.

2. autoMate's own web chat (the lightweight path)

Open the autoMate hub URL in a browser. Use the built-in chat tab. autoMate's own agent runs the loop — handy for quick queries without an external client.

Both modes share the same backend. Tools, files, notes, memory, and the agent loop are the same. Pick whichever entry point fits the moment.

Connecting in mode 1

Open autoMate Settings → Connect to AI clients and click "Copy install text (all clients)". You get a single markdown blob with the URL + token already filled in, plus per-client sections for OpenClaw, Claude Desktop, Cursor, Cline, and generic MCP clients.

Three ways to use that text:

• Read it yourself and edit your client's config file by hand. Each section tells you the file path and exact JSON to add.
• Paste it into another AI ("Cursor, here's autoMate, set it up for me") — the text is written so an AI can read the section that matches its own client and edit the right config file.
• For OpenClaw specifically, paste into your OpenClaw config under bundle-mcp — the section spells it out.

After the client picks up the new server, it'll see all autoMate's tools (search.find, notes.read, files.list, ...) plus a top-level automate tool that runs autoMate's own agent loop on demand.

The MCP endpoint

POST {hub}/mcp/                       # note trailing slash
Authorization: Bearer {token}
Content-Type:  application/json
Accept:        application/json, text/event-stream

Standard streamable-HTTP MCP — initialize, then tools/list and tools/call. autoMate exposes ~40 tools out of the box.

The token is the same Bearer token used by the v4.5.6 inbox endpoint. Stored in the existing settings KV table; rotate via Settings → Channels → Regenerate.

The legacy HTTP inbox (v4.5.6)

For tools that don't speak MCP — n8n, Zapier, custom Python / Node scripts, iOS Shortcuts — there's still a flat HTTP inbox:

POST /api/channels/inbox
Authorization: Bearer <token>

{"channel": "wechat", "user_id": "wxid_x", "text": "find my notes"}

→ {"text": "...", "run_id": "...", "ms": 1842}

Kept around because it's strictly easier than MCP for non-AI gateways. But for AI clients, MCP is cleaner.

What about v4.5.6's "OpenClaw outbound webhook" plan?

That was the wrong direction. The honest take:

• OpenClaw is an agent.
• autoMate is also an agent (with its own LLM, memory, etc.).
• Two agents fighting over one message stream is a bad pattern.
• Treating autoMate as a tool that OpenClaw's agent calls when needed is the clean pattern. That's what v4.5.7 ships.

The v4.5.6 inbox endpoint isn't deleted — it still works for cases where you genuinely want autoMate to be the only brain (a custom Telegram bot that just wraps autoMate, no OpenClaw). But it's no longer the highlighted path.

Deprecation status of automate/bots/

The old per-platform bot adapters (telegram.py, wechat_oa.py, wecom.py, wechat_personal.py) remain frozen. New deployments should use OpenClaw + the MCP bridge above. The old code will be removed once OpenClaw integration is proven in the wild.