---
type: "post"
id: "copilot-cli-1-0-112-counts-resumed-session-6b78c3"
slug: "copilot-cli-1-0-112-counts-resumed-session-6b78c3"
title: "Copilot CLI 1.0.112 counts resumed-session AI credits toward the session limit"
description: "GitHub Copilot CLI 1.0.112 makes AI credits already used in a session count when that session is resumed, and makes subagents honor the session-wide --max-ai-credits cap."
retrieval_nugget: "GitHub Copilot CLI 1.0.112 makes AI credits already used in a session count when that session is resumed, and makes subagents honor the session-wide --max-ai-credits cap. It also adopts each model’s recommended context window by default. For teams operating long-running coding tasks, the change makes session controls more consistent."
published_at: "2026-11-14"
updated_at: "2026-11-14"
record_date: "2026-11-14"
date_kind: "published_at"
topics: ["github"]
entities: ["GitHub","GitHub Copilot CLI"]
related_patterns: ["harness-architecture-outlives-model-choice"]
novelty_label: "new_fact"
what_changed: "Copilot CLI 1.0.112 changes credit accounting so AI credits used in resumed sessions count toward the session limit."
prior_refs: ["https://newruntime.com/posts/github-copilot-cli-v1-0-109-adds-claude-c94543/"]
source_url: "https://github.com/github/copilot-cli/releases/tag/v1.0.112"
source_urls: ["https://github.com/github/copilot-cli/releases/tag/v1.0.112"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:copilot-cli-1-0-112-counts-resumed-session-6b78c3"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/copilot-cli-1-0-112-counts-resumed-session-6b78c3/","markdown":"https://newruntime.com/posts/copilot-cli-1-0-112-counts-resumed-session-6b78c3.md","json":"https://newruntime.com/posts/copilot-cli-1-0-112-counts-resumed-session-6b78c3.json"}
markdown_url: "https://newruntime.com/posts/copilot-cli-1-0-112-counts-resumed-session-6b78c3.md"
---

# Copilot CLI 1.0.112 counts resumed-session AI credits toward the session limit

## Retrieval answer

GitHub Copilot CLI 1.0.112 makes AI credits already used in a session count when that session is resumed, and makes subagents honor the session-wide --max-ai-credits cap. It also adopts each model’s recommended context window by default. For teams operating long-running coding tasks, the change makes session controls more consistent.

## Summary

GitHub Copilot CLI 1.0.112 makes AI credits already used in a session count when that session is resumed, and makes subagents honor the session-wide `--max-ai-credits` cap. It also adopts each model’s recommended context window by default. For teams operating long-running coding tasks, the change makes session controls more consistent.

## What happened

GitHub’s release notes for Copilot CLI 1.0.112 are dated November 13, 2026. The release contains a long list of terminal, session, plugin, MCP, sandbox, and interface fixes, alongside several changes to how the coding agent applies session-level controls.

The central accounting change is explicit: previously used AI credits now count toward the session limit when a user resumes a session. A session that is paused and later reopened therefore retains the credits already consumed for purposes of its limit. The notes do not announce a new number of credits, a new plan, or a pricing change. They describe a change in the application of an existing session limit.

GitHub also says that subagents now respect the session-wide AI credit limit configured with `--max-ai-credits`. This matters because subagents can perform work within a broader coding task while the primary interaction remains in the parent session. Under the documented behavior, the configured session ceiling applies to that delegated work as well. The release notes do not specify how the CLI displays the allocation between a parent interaction and individual subagents, but they state that the cap is session-wide.

A separate default changes context handling. Copilot CLI now uses each model’s recommended context window by default, while preserving explicit context choices. In practical terms, a user who has selected a context window keeps that choice; where no explicit setting exists, the CLI selects the recommendation associated with the model. GitHub does not enumerate those recommended windows in these notes or explain the criteria used to set them.

The release also makes several session-related interface corrections. Session-limit dialogs stay attached to their own session when a user switches sessions, leaving other sessions available for typing. The final-model-call prompt no longer rounds used AI credits up to the limit: GitHub gives the example of displaying 299.6 of 300 rather than saying the limit has been reached. The `/tasks` command now includes subagent follow-up runtime after a resumed session.

Elsewhere, GitHub documents behavior around sandboxing and integrations. Its updated sandbox help says that on macOS, sandbox host rules block direct TCP connections to local ports; localhost remains reachable through the local proxy via HTTP or a CONNECT tunnel from a proxy-aware client. The release also updates `/chronicle cost-tips` to include Copilot App sessions, recommend cheaper models and reasoning settings based on usage, and report sooner when local usage is available. These are accompanying changes rather than evidence that the new credit accounting reduces spending.

## What the archive adds

New Runtime’s earlier coverage on [Copilot CLI model selection in long-context sessions](https://newruntime.com/posts/github-copilot-cli-1-0-105-lets-users-77ef51/) described version 1.0.105 restoring the ability to select Copilot Fusion High or Copilot Fusion Max during long-context sessions. This is a new release and a new fact on the Copilot CLI story: it concerns accounting and enforcement of session controls rather than model selection.

## Trend signal

This modestly supports the [harness architecture outlives model choice](https://newruntime.com/patterns/harness-architecture-outlives-model-choice/) pattern. The selected model remains relevant, but the CLI’s session accounting, delegation boundaries, context defaults, and interface behavior determine how that model is used in production work. The release is evidence of a harness-level adjustment, not evidence that one model is superior to another.

## Why now

GitHub dated the release November 13, 2026. The notes identify the shipped behavior but do not explain the timing, connect it to a new commercial policy, or describe a triggering customer problem. It should therefore be read as a concrete implementation update rather than a stated strategic shift.

## Why it matters

A credit cap is operationally useful only when it follows the unit of work a team is trying to govern. Long coding tasks do not always complete in one uninterrupted terminal session: developers pause work, return later, switch sessions, and delegate investigation or implementation to subagents. Counting consumed credits after resumption means the configured ceiling applies to that continuing task rather than only to its most recent active interval.

Subagent enforcement is equally important for predictability. A team may set `--max-ai-credits` to constrain an exploratory task, a repair run, or a broad code review. If delegated work is subject to the same session-wide setting, the cap becomes a clearer guardrail for the entire task. That does not make the limit a complete cost-control system: GitHub has not published how credit use maps to money, task duration, or output quality.

The context-window default is a smaller but related operational decision. Model-specific recommendations can remove one manual choice for users who have not tuned context settings, while explicit settings remain available for teams that need predictable configuration. Together, the changes make the CLI’s behavior more legible, but teams should validate it against their own models, tasks, and agent configurations.

## Limits and uncertainty

GitHub’s release notes do not quantify typical credit consumption in resumed sessions, measured overruns, cost savings, or the frequency with which developers hit session limits. They do not give the numerical default limit, explain credit accounting at the level of individual parent and subagent actions, or list the recommended context window for every model. The notes also do not state whether these changes affect every authentication, billing, or model-provider configuration in the same way. New Runtime has not independently tested the release. Teams should confirm the behavior in their installed version and in the workflows that matter to them.

## Skill to build now

Create a small acceptance test for your agent budget policy this week. Start a Copilot CLI session with a deliberately low `--max-ai-credits` value, consume part of it, resume the session, and observe the displayed usage and limit behavior. Repeat the test with a subagent enabled. Record the CLI version, model, context setting, task type, and observed result so that upgrades can be checked against the same scenario.

## Reader takeaway

Copilot CLI 1.0.112 makes an existing session credit cap more consistent across resumed work and subagent activity, while defaulting to model-recommended context windows when no explicit choice exists. If your team uses session limits as a guardrail, test a resumed multi-agent task after updating rather than assuming the cap behaves as before.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Github](https://newruntime.com/topics/github.md): Explore the github topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)