Radar brief — 2026-09-13

What the evidence shows

Fifteen curated signals across 23 tracked assistants. About half the mix is GitHub/changelog-style; the rest is HTML/docs, so treat Windsurf, Kiro, Tabnine, and Factory as noisier than the versioned GitHub tags. One source in this window reported errors. Several releases omit dozens of extra notes (bodyOmitted) — those omitted lines are not inferred below.

Versioned GitHub releases (preferred)

HTML / docs changelogs (lower confidence)

For developers

• Score plugins, don’t eyeball them. On Claude Code v2.1.269 run claude plugin eval --help and keep the JSON/HTML report next to any plugin you ship or pin. After Codex rust-v0.153.0 / v0.154.0, list/install/remove marketplace plugins from the CLI and expect existing sessions to pick up new plugin tools, skills, and hooks after upgrades or rollbacks.
• Isolate agent checkouts. Codex --worktree / /worktree is still experimental — use it for forked sessions, then browse/resume those checkouts instead of letting the agent mutate your primary tree.
• Turn on the diffs you already paid for. Enable Claude bashEditDiffEnabled so Bash file edits return a diff in the tool result. In OpenCode v1.18.26, treat empty apply_patch move paths as a permission-metadata bug that is now fixed — re-check any hook that keyed off that field.
• If you use Cline Desktop, install gh. The composer PR/CI row only appears with GitHub CLI signed in and a GitHub.com origin; it will not push or open the PR for you. See Install gh CLI so Cline composer shows live PR and CI status.
• Stop assuming empty model catalogs mean “no tools.” Cline cli-v3.0.61 restores tool calling (and image reads) for Dify, SAP AI Core, opencode, and Codex CLI models whose catalogs declared no capabilities. If those models looked “chat-only” last week, re-test.
• Fix resume and MCP flake before blaming the model. Antigravity --continue now falls back to the latest non-empty workspace conversation. Cline CLI gives unreachable SSE/HTTP MCP a 10s budget instead of tearing down the TUI. OpenCode Claude 5 sessions no longer die on stale thinking blocks.
• Kiro V3: inspect before you copy. Use /sessions and /config, then decide whether cloud Steering/Skills/Hooks should load locally. See Enable Apply your cloud configuration to local sessions in Kiro V3.
• Windsurf users: update muscle memory. ACP is always on; Cascade-era new-agent, mentions, Deploy, Vibe & Replace, and workflow commands are gone. Prefer saved SSH hosts in the location selector.

For teams

• Treat GPT-6-Astra and worktrees as a paired rollout. Codex rust-v0.154.0 puts GPT-6-Astra in the picker and Bedrock catalogs, and adds isolated checkouts. Default that pairing rather than turning the model on against a shared working tree: Switch Codex CLI to GPT-6-Astra and isolated --worktree checkouts.
• Do not silently run mixed CLI/Hub generations. Cline cli-v3.0.61 will ask before replacing an older Hub and will drain in-flight turns; a wedged Hub is left alone. Put that prompt in runbooks so CI/one-shot jobs are not left on stale code. Windows binaries are now Authenticode-signed — expect application-control policies to start failing closed with a printed error instead of a bare crash.
• Cap fan-out and tag telemetry. Claude Code’s workflow concurrent-agent limit is now 1–256 (CLAUDE_CODE_WORKFLOW_MAX_CONCURRENT_AGENTS). Set it on purpose. Turn on OTEL_METRICS_INCLUDE_REPOSITORY if you need vcs.* attributes on metrics/events.
• Rehearse MCP OAuth and schema changes. Codex coordinates token refresh and surfaces login challenges without replaying rejected tools. Antigravity now preserves open object schemas (additionalProperties: true) instead of rejecting extra arguments. Goose v1.48.0 adds on_failure PreToolUse hooks and a stable tool_call_id — wire those into existing permission/hook pipelines before the provider list grows again.
• Goose provider sprawl is a routing problem, not a trophy. v1.48.0 adds several declarative OpenAI-compatible providers, GLM-5.3, custom cost fields, and Bedrock GPT-5.6. Pin providers and cost tracking in config; do not assume a later Goose cut from proposals (v1.50.0 is not in this evidence).
• Windsurf breaking changes are team-visible. Always-on ACP, removed Cascade entry points, SSH remote hosts, and cloud network allowlist cards that can update a security profile belong in the team changelog: Add saved SSH agent hosts in Windsurf and drop Cascade commands.
• Zed v1.19.0-pre is a prerelease. Useful for memory/streaming on long agent sessions; do not make it the org default from this tag alone.
• Ignore Factory’s 60% GPT/Grok discount as a platform decision. It is a limited-time selector badge on a low-score docs changelog, not a capability change.

What to ignore

• Omitted changelog tails. Claude Code (+97), Goose (+98), Codex 0.154.0 (+147) and 0.153.0 (+95), Zed (+50), Cline Desktop (+36) — do not invent those lines.
• Duplicate / mixed HTML pages. Kiro 2.21.3 repeats the 2.21.4 session-dashboard story. Tabnine’s 6.5.4 page still carries 6.5.3–6.5.1 notes. Factory Droid is a docs changelog at the bottom of the scoreboard.
• Proposals that are not releases. Open items mention Goose v1.50.0, a Cline switch to “Claude Fable 5.1,” and Grok CLI prompt/MCP cleanups. None of those versions/facts are in the evidence JSON; do not treat them as shipped.
• Zed editor chrome as agent news. Call hierarchy and Git Panel multi-select are real, but they are not model/agent regressions.

Watch next

• Whether Codex worktrees leave “experimental” and whether GPT-6-Astra becomes the default rather than a picker entry.
• Whether claude plugin eval reports start showing up in plugin READMEs (that is the only reproducible scoring surface in this window).
• Goose’s declarative-provider list vs. actual default routing; the open team proposal to move past v1.48.0 is not confirmed here.
• Cline Hub replacement behavior in CI, and whether empty model capability lists appear for other providers.
• Windsurf fallout from deleted Cascade commands and always-on ACP.
• Eight open improvement proposals, including Grok CLI prompt/MCP cleanup items that have no matching release in this brief.

How to use DevAgentRadar

This brief is a public, evidence-only cut: 15 signals, 23 assistants tracked, 49% GitHub/changelog mix, 8 open proposals, 1 source reporting errors. Upgrade from the versioned GitHub rows first; use HTML/docs rows as leads, not proof. If a change was omitted from a release body, it is omitted here on purpose. Developers should take the concrete CLI flags and local setup steps; teams should take the permission, MCP, Hub, and breaking-command items. Follow an open proposal only after it matches a version you can actually install.

Evidence appendix

• Evidence window: top curated signals (versioned releases preferred)
• Assistants tracked: 23
• Signals in this brief: 15
• Trusted source mix (GitHub/changelog vs HTML): 49%
• Open improvement proposals: 8
• Sources reporting errors: 1

High-signal releases used

• Claude Code v2.1.269: Added claude plugin eval: run a plugin's eval suite against Claude Code and get scored, reproducible results (JSON + HTML report); see `claude plugin eval --h — source

• Goose v1.48.0: TrustedRouter declarative provider #11422 · OpenCode Zen gateway as a declarative provider [#10939](https://gi — source

• OpenAI Codex CLI rust-v0.154.0: GPT-6-Astra is now available in the model picker and Amazon Bedrock catalogs. (#42879, #42619) · Experimental worktree support lets you create isolated checkout — source

• Zed AI v1.19.0-pre: This week's release includes call hierarchy support, multi-select in the Git Panel, automatic language detection for untitled buffers, and project search on typ — source

• Cline desktop-v0.0.26: The composer now shows the current branch's GitHub pull request — PR number, merge status, changed-line totals, and CI checks. · Click through to open it in you — source

• Google Antigravity 1.2.1: Added support for excludeDefaultComponents: true in custom agent Markdown frontmatter, allowing custom agents to opt out of default prompt sections and built- — source

• OpenAI Codex CLI rust-v0.153.0: Vim mode now supports undo with u and redo with Ctrl+R, preserving complete drafts including pasted content and attachments. ([#41941](https://github.com/op — source

• Cline cli-v3.0.61: Cline now handles a running Hub that is older than your CLI. · Instead of quietly talking to a hub executing stale code, you get a prompt showing how many activ — source

• OpenCode v1.18.26: Claude 5 sessions now tolerate stale thinking blocks instead of failing after prompt or tool changes. · Bedrock GPT-5.6 models now accept none reasoning effor — source

• Windsurf v3.10.23: v3.10.23 September 10, 2026 Devin Desktop Remote agent hosts: connect to agents over SSH from the location selector, with saved hosts, status dots, and a server — source

• Kiro v2.21.4: 2.21.4 Sep 11, 2026 · 2.21.3 Sep 10, 2026 · 2.21.2 Sep 8, 2026 +1 more Session Dashboard, Configuration Panel, and Cloud Sync Browse and manage local and cloud — source

• Tabnine v6.5.4: v6.5.4 August 30, 2026 Bug Fixes Tabnine CLI: Provenance & Attribution : Fixed attribution feedback being dropped after tool inputs were edited via VS Code diff — source

• Kiro 2.21.3 — Sep 10, 2026: September 1, 2026 Session Dashboard, Configuration Panel, and Cloud Sync · 2.21.0 · Browse and manage local and cloud sessions, inspect configured resources, an — source

• Factory Droid v0.209.0 — · Desktop v0.166.0 (September 1, 2026): Session archiving from chat, a new default model, faster startup, and steadier app views · Archive sessions from chat - Archive the current session with /ar — source

Open proposals

• Install gh CLI so Cline composer shows live PR and CI status (audience: developer, P5)
• Add saved SSH agent hosts in Windsurf and drop Cascade commands (audience: team, P5)
• Upgrade Goose to v1.50.0 and enable GPT-6 Astra with goose-agent tool calling (audience: team, P5)
• Switch Codex CLI to GPT-6-Astra and isolated --worktree checkouts (audience: everyone, P5)
• Enable Apply your cloud configuration to local sessions in Kiro V3 (audience: developer, P4)
• Pin Cline models after the v3.0.61 switch to Claude Fable 5.1 (audience: everyone, P4)
• Replace <output_efficiency> with <response_guidelines> in Grok CLI prompts (audience: developer, P4)
• Drop legacy managed MCP configs and shadowed platform skills in Grok CLI (audience: developer, P4)