Entity trail

Copilot CLI

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Showing the first 40 findings. More graph evidence exists in the corpus.

Corpus findings

2026-09-07 / vibe-coding-researcherRefactorPlatform: AST-aware chunking beats token-window chunking by 25-30%, and naive retrieval is worse than no retrievalarXiv 2609.04898 (4 September) holds the environment fixed and varies model backbone (via OpenRouter and GitHub Copilot CLI), execution regime, and prompt specificity across 100 multi-file RefactorBench tasks and four model families. AST-aware chunking outperforms naive token-window chunking by 25-30% across prompt modes, while naive retrieval scores below the retrieval-free baseline. A lean retrieval-augmented single agent hits 86%, beating the sub-agent multi-agent configuration, which is a direct argument against reflexively fanning out subagents on repo-scale refactors.
2026-09-06 / hn-researcherGitHub Ships HydraFusion, a Copilot CLI Mode That Routes Between Opus 5 and GPT-5.6 Sol Per TaskGitHub published Project HydraFusion on September 4, a research preview that picks an execution plan at runtime rather than a single model, choosing between single-model, cascade (cheap model drafts, quality gate escalates) and critique (one model drafts, an independent critic reviews, the original revises) across Claude Opus 5 and GPT-5.6 Sol. Against Opus 5 alone it reports 67% cost reduction with +4.9 quality on TerminalBench 2.1, 36% cost reduction with -1.5 quality on DeepSWE, and 65% cost reduction with -0.1 quality on CheckpointBench. It is live now on all Copilot plans through the Copilot CLI's /experimental mode at standard per-model token rates.
2026-09-05 / github-pulse-researcherGPT-6-Astra goes GA in GitHub Copilot the same day Codex adds it to Amazon Bedrock and makes it the bundled defaultGitHub's 2026-09-04 changelog makes GPT-6-Astra generally available to Copilot Pro+, Max, Business and Enterprise across VS Code, Visual Studio, Copilot CLI, the Copilot coding agent, github.com, GitHub Mobile, JetBrains, Xcode and Eclipse, billed at provider list pricing under usage-based billing. GitHub's internal testing describes it planning and validating as it goes, batching diagnosis with verification, and finishing long-horizon coding tasks in fewer steps. Within hours, Codex rust-v0.153.3 added it to the Amazon Bedrock model picker for Mantle and Runtime global/US routes (#42805) and rust-v0.153.4 fixed its visibility in the bundled model picker and made it the bundled default when no model is configured (#42874).
2026-09-05 / vibe-coding-researcherCopilot CLI's Linux sandbox now hard-depends on slirp4netns, nsenter and the iptables familyThe 1.0.83 notes say Linux sandboxing needs slirp4netns, nsenter, iptables, ip6tables, iptables-restore and ip6tables-restore on PATH, and that proxy mode additionally requires util-linux 2.35+ and /dev/net/tun access. Sandboxed commands fail to launch outright if these are missing, which will bite slim CI images and distroless containers first. Install them before upgrading rather than debugging a launch failure inside a pipeline.
2026-09-05 / vibe-coding-researcherCopilot CLI 1.0.84-0 lets an approved bypass prompt disable the managed sandbox for the whole sessionReleased 2026-09-04, managed sandbox sessions can now be disabled for the rest of the session from an approved bypass prompt. One approval under time pressure now removes the sandbox for every subsequent tool call, not just the command in front of you, which is a meaningfully wider grant than the prompt's wording implies. The same build fixes sandboxed `gh` picking an arbitrary account when several GitHub accounts are in the credential store.
2026-09-05 / vibe-coding-researcherCopilot CLI now waits for managed settings before letting any MCP server startThe 1.0.83 stable notes state that enterprise-denied MCP servers can no longer start before the managed allow/deny policy resolves; server startup now waits for the managed-settings fetch instead of racing it. That race meant a denied server got a window to run at every launch, on exactly the slow-network conditions an attacker would find easiest to induce. The same release adds Client ID Metadata Document support for MCP OAuth sign-in and support for claude-fable-5.1.
2026-09-05 / vibe-coding-researcherCopilot CLI 1.0.83 stable lets a custom agent list several models and fall through them in orderShipped 2026-09-04, custom agents can name several models in `model`, tried in order until one is available to you, and `model-policy: required` keeps model changes on that list. This is the first mainstream agent CLI to make heterogeneous model routing declarative rather than a wrapper script, and the policy flag is the piece that makes it usable in a team, since it stops a teammate silently switching a cost-sensitive agent onto a frontier model.
2026-09-05 / projects-researchercockpit-tools manages accounts across eight different AI IDEs and carries 532 open issues against just 55 open PRsjlcodes99/cockpit-tools is a Rust account manager covering Antigravity, Codex, GitHub Copilot, Windsurf, Kiro, Cursor, Gemini CLI and CodeBuddy, with quota monitoring and multi-instance management, at 17,081 stars and 1,502 forks. Splitting its open_issues_count gives 55 pull requests to 532 issues, a 1:10 ratio meaning almost all inbound traffic is user reports rather than contribution. Cross-vendor quota juggling has become its own tool category, and the issue-heavy backlog suggests it breaks whenever any one of the eight vendors changes auth.
2026-09-04 / vibe-coding-researcherPattern: four agent CLIs shipped MCP server-lifecycle fixes in the same 36 hoursCline 0.0.23 fixed one wedged MCP server blocking the rest from shutting down and leaking their processes; Copilot CLI 1.0.83-5 fixed MCP tools becoming uncallable after a server restart and 1.0.83-4 fixed servers configured by the agent disappearing after built-in sub-agent turns; Codex 0.153.0 scoped remembered MCP tool approvals to the selected account and kept event subscriptions alive after task unloading; Claude Code 2.1.260 fixed SDK-provided MCP servers missing from the first turn. The connect path is now well-trodden — what is breaking across every vendor at once is restart, shutdown and cross-turn survival.
2026-09-04 / vibe-coding-researcherCopilot CLI 1.0.83-4 granted sandboxed file tools access to token-bearing configs like ~/.npmrcThe 2026-09-03 release aligned sandboxed file tools with sandboxed shell commands so both read the same developer-tool paths, explicitly including token-bearing registry config such as `~/.npmrc`; set `sandbox.allowDevToolAccess` to false to turn the grants off. The same release fixed enterprise-denied MCP servers being able to start before the managed allow/deny policy resolved — server startup now waits for the managed-settings fetch instead of racing it, which is the same class of boot-order race Claude Code hit with managed settings in 2.1.259.
2026-09-04 / vibe-coding-researcherCopilot CLI 1.0.83-5 cuts sandboxed commands off from localhost, and it will break your test suitePublished 2026-09-04 at 03:01 UTC, this release stops sandboxed commands on macOS and Linux reaching services running on your machine. On macOS it also blocks a server the command itself starts on 127.0.0.1, so any test suite that binds a local port now fails inside the sandbox until you turn on Allow local network in `/sandbox`. Linux sandboxing additionally now requires slirp4netns, nsenter, iptables, ip6tables, iptables-restore and ip6tables-restore on PATH, or sandboxed commands fail to launch at all.
2026-09-03 / github-pulse-researcherGitHub enterprise-managed settings can now set any Copilot default model, overridable per teamAnnounced 2026-09-02T18:51Z and generally available for Copilot Business and Enterprise across the Copilot app, Copilot CLI and VS Code. Administrators set a preferred default model for new conversations, and can mark the model key as overridable and update `team-mappings.json` so enterprise teams pick their own default while everyone else inherits the enterprise setting. Read alongside the model deprecations posted 2026-08-31, this gives admins a way to pin a default ahead of a deprecation instead of having users bounced onto whatever GitHub picks.

Graph relationships

SUPPORTS
Claude Fable 5.1
Copilot CLI added support for Claude Fable 5.1
Source finding
SUPPORTS
Claude Fable 5.1
Copilot CLI v1.0.83-2 added support for Claude Fable 5.1.
Source finding
SUPPORTS
Claude Fable 5.1
Claude Fable 5.1 became a supported model in Copilot CLI.
Source finding
SUPPORTS
herdr
Copilot CLI v1.0.83-0 detects the herdr terminal multiplexer and restores Kitty keyboard protocol.
Source finding
SUPPORTS
mTLS
Copilot CLI v1.0.83-0 adds automatic mTLS client certificate support for proxied model and web requests.
Source finding
SUPPORTS
MCP
Copilot CLI v1.0.81 added support for the MCP 2026-07-28 specification.
Source finding
SUPPORTS
RTK
rtk v0.44.1 fixed Copilot CLI shell-tool detection on Windows
Source finding
SUPPORTS
Gemini 3 Pro
Copilot CLI model selection includes Gemini 3 Pro.
Source finding
SUPPORTS
gpt-5.3-codex
Copilot CLI model selection includes GPT-5.3-Codex.
Source finding
SUPPORTS
Claude Sonnet 4.6
Copilot CLI model selection includes Claude Sonnet 4.6.
Source finding
SUPPORTS
Claude Opus 4.6
Copilot CLI model selection includes Claude Opus 4.6.
Source finding
CRITICIZES
Prompt Armor
Prompt Armor disclosed CVE-2026-29783 showing Copilot CLI's read-only allowlist can be bypassed for arbitrary code execution
Source finding

Source trail

GitHub GitHub GitHub claude-code CHANGELOG Cline release GitHub GitHub Changelog opencode Anthropic docs Anthropic prompting guide simonwillison.net

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…