Definition
Push delivery is sending the result of work an AI agent asked for back into that agent's chat as a new message, once the chat can take one, instead of waiting for the agent to collect it.
Also called: result delivery, pushed results.
An agent that starts slow work (a review, a helper chat, a workflow run) can wait for it, poll for it, or end its turn. Waiting and polling keep the chat busy; ending the turn means the result has to reach it some other way. Push delivery is that other way: the result arrives as a message, and the agent picks up from there.
The hard part is timing. A message sent into a chat mid-turn can interrupt it, so a good push waits until the chat is idle, or goes into a queue the app works through when it's free.
In Pairbar
Deliver results into the chat that asked is on to start with (Settings → Integrations → Agent actions). A finished helper's, follow-up's or workflow run's result that the chat didn't collect with wait or get_result is sent into it:
- not before 90 seconds after it finished, so a chat calling
waitgets it that way instead; - Codex: submitted through
codex queue; a receipt confirms queue submission, not that the desktop has started processing it; - Claude: only while Claude Code's own record says the chat is idle, never mid-turn or while it waits for you; a chat busy for 30 minutes gets nothing pushed;
- held while agent actions are paused or Pairbar is locked, tried up to three times, and kept across a quit.
The message says which request it answers, then the result framed as another chat's information, not instructions. Agent activity shows a delivered row. See Results.
Frequently asked questions
- Will a pushed result interrupt my chat?
- Codex receives a queue submission rather than a new model turn started by Pairbar; queue receipt does not prove processing. Claude receives the result only while its chat is idle.
- Can I turn it off?
- Yes. Turn off Deliver results into the chat that asked, and results are only ever collected by the chat with
waitorget_result. - Which results are pushed?
- Only a finished helper's, a follow-up's or a workflow run's, for a Claude or Codex chat Pairbar identified, and never one the chat already collected.
Related
- DocsIntegrationsLive status for Claude and Codex through their hooks, Pairbar's tools for their chats (the plugin, the `pairbar` command, the tool reference, who's calling, agent activity), and agent actions: second opinions, reviews, helper chats, handoffs, workflow runs and drafts, and check-ins that agents ask for and you approve, with budgets and guardrails
- GlossaryApproval ticketThe record of one agent's request: what it asked for, the decision, the work it started and the result the agent collects.
- Use caseSend a follow-up to a Claude Code or Codex chat without opening it⋯ → Send message… (M) confirms a Codex queue submission with an Open in Codex link, or continues a Claude session in the background once it's idle.
- Use caseGet a second opinion from Codex without leaving your Claude Code chatA Claude chat asks Pairbar for a second opinion; after you approve, a read-only Codex chat answers and the reply comes back into the Claude chat.