Skip to main content
How OpenClaw delivers replies into Matrix rooms: streaming previews, inbound voice notes, and reactions.

Streaming previews

Matrix reply streaming is opt-in. streaming.mode controls how OpenClaw delivers the in-flight assistant reply; streaming.block.enabled controls whether each completed block is kept as its own Matrix message.
To keep live answer previews but hide interim tool/progress lines:
The full config accepts { mode, chunkMode, block, preview, progress }:
  • progress.label: custom label, "auto"/unset to pick a configured or built-in label, or false to hide it.
  • progress.labels: candidates used only when label is "auto" or unset.
  • progress.maxLines: max rolling progress lines kept in the draft; older lines are trimmed past this.
  • progress.maxLineChars: max characters per compact progress line before truncation.
  • progress.toolProgress: when true, live tool/progress activity appears in the draft. The default false keeps the draft to its headline, commentary, plan milestones, and approval or failure lines.
streaming.block.enabled (default false) is independent of streaming.mode: Notes:
  • If a preview grows past Matrix’s per-event size limit, OpenClaw stops preview streaming and falls back to final-only delivery.
  • Media replies always send attachments normally. If a visible preview cannot be reused safely, OpenClaw keeps it until the complete replacement is confirmed and then redacts it. If replacement delivery fails, is partial, or produces no visible event, the preview remains visible.
  • Tool-progress preview updates are on by default when preview streaming is active. Set streaming.preview.toolProgress: false to keep preview edits for answer text but leave tool progress on the normal delivery path.
  • Preview edits cost extra Matrix API calls. Leave streaming.mode: "off" for the most conservative rate-limit profile.
  • Legacy scalar/boolean streaming values and the flat blockStreaming / chunkMode keys are rewritten to this nested shape by openclaw doctor --fix.

Voice messages

Inbound Matrix voice notes are transcribed before the room mention gate, so a voice note saying the bot name can trigger the agent in a requireMention: true room, and the agent gets the transcript instead of only an audio attachment placeholder. Matrix uses the shared audio media provider under tools.media.audio, such as OpenAI gpt-4o-mini-transcribe. See Media tools overview for provider setup and limits.
  • m.audio events and m.file events with an audio/* MIME type are eligible.
  • In encrypted rooms, OpenClaw decrypts the attachment through the existing Matrix media path before transcription.
  • The transcript is marked machine-generated and untrusted in the agent prompt.
  • The attachment is marked as already transcribed so downstream media tools do not transcribe it again.
  • Set tools.media.audio.enabled: false to disable audio transcription globally.

Reactions

Matrix supports outbound reactions, inbound reaction notifications, and ack reactions. Outbound reaction tooling is gated by channels.matrix.actions.reactions:
  • react adds a reaction to a Matrix event.
  • reactions lists the current reaction summary for a Matrix event.
  • emoji-list discovers custom emoji from the current conversation’s room packs and your personal pack.
  • emoji="" removes the bot’s own reactions on that event.
  • remove: true removes only the specified emoji reaction from the bot.
emoji-list reads MSC2545 im.ponies.room_emotes packs from the authorized current room and im.ponies.user_emotes account data. It returns up to 100 sorted entries such as { "name": "party", "identifier": "party", "url": "mxc://example.org/party" }; sticker-only entries are excluded. Pass identifier to react: it is the plain shortcode stored directly as the Matrix reaction’s m.relates_to.key, not the mxc:// media URL. Custom-reaction rendering depends on the Matrix client, so url is included separately for clients or agents that need the image. Resolution order (first defined value wins): reactionNotifications: "own" forwards added m.reaction events when they target bot-authored Matrix messages; "off" disables reaction system events. Reaction removals are not synthesized into system events - Matrix surfaces those as redactions, not as standalone m.reaction removals.