All stories
Vespin Ecosystem Community reports

Your Slack thread can wake a coding agent again in Vespin 0.25.1

The community coding-agent inbox now revives archived agents from Slack replies, continues stopped work in a new agent and carries pictures in both directions.

By AITrending2 min read
Sources (4)

Why now The November 3 release of version 0.25.1 adds concrete Slack follow-up behavior and a specific boundary on offering Mac commands after other participants contribute.
Vespin inbox with a coding-agent thread asking for approval of a proposed direction.
Author-supplied Vespin inbox illustration. It shows the general task-and-decision interface, not the new Slack continuation workflow; task labels are not independently verified outcomes. Source: Vespin / ReviewStage

A follow-up in Slack can bring a coding agent back to work. Vespin 0.25.1, released November 3, says a message in a Slack thread revives its archived agent—or continues the work in a new agent if the previous one has stopped. When a Slack turn finishes, the agent archives itself.

That makes the thread a more useful place for a continuing task: send another instruction, get the answer back there, and pass pictures along with the conversation. The release adds images in both directions through Vespin’s bot. It also fixes follow-ups that previously produced only “The agent finished.” instead of the agent’s answer, and replies to Slack DMs that could start a stray agent.

A thread with a visible trail

According to the tagged changelog, Slack-started agents take their names from the initiating message, carry a Slack mark in the app and show the original message as a linked transcript line. The bot reacts with an eyes emoji when it acts on a message, replacing an “On it” reply; its messages omit link previews.

There is also a precise permission-related change: a Slack-started agent stops offering to run commands on the user’s Mac once someone else’s words from the thread reach it. That is the project’s stated rule, not evidence of a universal prompt-injection defense.

Vespin is the recently renamed Luke: an inbox and task workspace around existing Claude Code or Codex agents, not a new coding model. The README offers a web app or an Apple silicon Mac download requiring macOS 14 or later. Setup involves signing in and connecting GitHub under Settings › Workspace; agents use the user’s own Claude or ChatGPT subscription or API key.

The repository is Apache 2.0, while the proprietary Claude Agent SDK used by hosted agents is outside that license. Cloud agents run in service sandboxes, so Slack integration should not be read as a promise that all conversation and execution stay on a Mac. These changes are reported from the project’s English release and documentation; AITrending has not independently reproduced the workflow.

Related reading