Messaging
PwrAgent’s messaging surface lets you drive Codex threads from the
chat platforms you already use — Telegram, Discord, Slack,
Mattermost, Feishu / Lark, LINE. Everything below is operator-facing;
the contributor / architecture content lives in the main repo under
docs/messaging-*.md.
Where to start
If you’re new to PwrAgent’s messaging story, the recommended path is Using Codex via Messaging → Providers:
- Using Codex via Messaging — the end-to-end usage guide. Bound threads, slash commands, the resume browser, the start card, debounce / queue / steer, monitor cards, detach. Same across every provider; per-provider quirks called out inline.
- Providers — per-platform setup walkthroughs. “What you need to get started” → “Step by step” → “Settings reference” for each of the six platforms.
Setting up providers
Pick your platform. Setup pages each follow the same structure (credentials needed → exact paste/save/test/pair flow → field-by- field Settings reference). Click the platform name to jump straight to its setup page.
| Platform | Inbound transport | Public port? | Setup |
|---|---|---|---|
| Telegram | Long polling | ✅ No | telegram |
| Discord | Gateway WebSocket | ✅ No | discord |
| Slack | Socket Mode | ✅ No | slack |
| Feishu / Lark | Persistent SDK WebSocket | ✅ No | feishu |
| Mattermost | HTTP callback (your host) | ❌ Yes | mattermost |
| LINE | HTTP webhook (your host) | ❌ Yes | line |
✅ No = PwrAgent dials out to the platform; nothing on your machine accepts incoming public traffic. ✅ is the safer default.
❌ Yes = the platform dials into a callback URL you host, usually fronted by Cloudflare Tunnel or Tailscale Funnel. Read Webhooks — a security note before standing one up.
Pairing
The Pairing flow is how you populate every platform’s authorized-user list (and shared-space allowlists) without having to find a numeric platform ID anywhere. Same mechanic across all six providers; the captured walkthrough plus the Messaging Activity troubleshooting screen for blocked inbound messages live at Messaging → Pairing.
Read before you toggle
A handful of cross-cutting reference pages worth reading once before configuring anything:
- Streaming responses — why the toggle is off by default and why turning it on usually makes things worse, not better.
- Webhooks — a security note — what’s at stake for the two HTTP-callback platforms (Mattermost, LINE), what you need to do at the tunnel layer, and when Mattermost can stay on a private network.
- Rate limits and budgets — per-platform write budgets from May 2026 PwrAgent probes, Slow Mode and Cool Off priority order, per-platform label caps.