Skip to content
Guidance/Daily Briefing
digesteveryonedigestevidenceradarmodelagent7 min read · Aug 9, 2026

Radar brief — 2026-08-09

Self-hosted runners, portable agent plugins, and safer auto-approvals dominate this window — Claude Code, Codex CLI, and Cline lead versioned GitHub signals while security defaults tighten.

What the evidence shows

Curated window across 22 tracked assistants (16 signals; ~55% GitHub/changelog-weighted). Highest-signal items below; prerelease and HTML changelog rows called out where noisier.

Claude Codev2.1.224
Key Signal & Impact
Self-hosted runner for web/mobile/desktop sessions (Team/Enterprise); archive plugin source over HTTPS with optional SHA-256 pinning; Bedrock region-prefix env; cancel-and-confirm on paste-affecting removes
OpenAI Codex CLIrust-v0.147.0
Key Signal & Impact
Portable Agent Plugins + multi-catalog search; conversation sections; --approve-for-me auto-reviewed approvals; import/sync Cursor & Claude skills/conversations without duplicates
OpenAI Codex CLIrust-v0.146.0
Key Signal & Impact
Named sessions, pin/switch side conversations; Agent Plugins manifests + Bedrock/Claude Code marketplaces; thread forks (incl. temporary); app-server ↔ remote Code Mode over WebSocket
OpenAI Codex CLIrust-v0.146.1
Key Signal & Impact
Safer automatic-review defaults for cyber-capable models; permission-change explanations in TUI
Clinedesktop-v0.0.8
Key Signal & Impact
Edit-any-earlier-message forks session, rewinds workspace checkpoint, re-runs; transactional/workspace-atomic restores; more reliable checkpoints after restart/compaction/resume
Clinev4.1.7
Key Signal & Impact
“View Changes” on completion rows via SDK checkpoints; pre-registered OAuth clients for remote MCP; centralized plugin settings with atomic toggles
Clinesdk/sdk/v0.0.72
Key Signal & Impact
Queued prompts survive aborts; durable session context across aborts/hub restarts; failed queued turns report run.failed
Google Antigravity1.1.10
Key Signal & Impact
Business sign-in for Gemini Enterprise (Cloud project, license seats, regional inference, admin controls)
Goosev1.45.0
Key Signal & Impact
Stable agent event IDs; model command tab-complete + provider switch; latest Gemini models; air-gapped GOOSE_DOCS_ROOT; templated compaction summaries
OpenCodev1.18.5
Key Signal & Impact
Claude adaptive-thinking fixes; OpenAI Responses phase fix; Mistral reasoning/cache stability; correct per-SDK prompt-cache keys; server terminal/review/discovery updates
Zed AIv1.15.0-pre
Key Signal & Impact
git.diff_base for gutter/status/diff vs HEAD or default-branch merge base; drag-out from Project Panel (macOS/Linux Wayland); JSX/TSX linked editing + Emmet in bodies (prerelease)
Gemini CLIv0.54.0
Key Signal & Impact
Session ID rotate on model fallback; HTTPS enforced for GoogleCredentialsAuthProvider (cleartext leak fix); caretaker triage hardening
Gemini CLIv0.56.0-nightly… / v0.54.0-preview.0
Key Signal & Impact
Caretaker triage/evals/Cloud Run workflows; capacity exhaustion as terminal error (nightly/preview — noisy)
xAI Grok CLI (Grok Build)8a14c91
Key Signal & Impact
Worktree flag across cwd/resume; signal-safe memory trace; skill-path suggestions; bounded .envrc eval; tools-server OOM protection (commit sync, not semver tag)
Windsurfv3.6.27
Key Signal & Impact
Windows cert-store load for Devin Desktop behind TLS-inspecting proxies; local edit/write/apply_patch/notebook_edit refuse symlink writes (HTML changelog)

For developers

  • Codex CLI (rust-v0.147.0 / 0.146.x): Try Agent Plugins catalogs and skill import from Cursor/Claude; use --approve-for-me only after reading the v0.146.1 safer cyber-model review defaults. Prefer pinned plugin sources over ad-hoc installs.
  • Claude Code (v2.1.224): If you need always-on agent compute, evaluate claude self-hosted-runner on your own machines/containers; pin archive plugins with SHA-256 when skipping git/npm.
  • Cline (desktop v0.0.8 + app v4.1.7 + SDK v0.0.72): Use edit-earlier-message + checkpoint rewind for safe prompt iteration; rely on “View Changes” for task diffs; expect queued prompts to survive aborts instead of vanishing.
  • Goose (v1.45.0): Use model-command tab completion / provider switch; set GOOSE_DOCS_ROOT for air-gapped docs; adopt structured compaction summaries if you compact often.
  • OpenCode (v1.18.5): Pull if you hit Claude thinking-shape breaks, OpenAI Responses phase issues, or Mistral cache/history drift.
  • Gemini CLI (v0.54.0): Stable tag fixes model-fallback session stickiness and forces HTTPS on Google creds — prefer over nightlies for daily use.
  • Grok Build (8a14c91): Note worktree/resume and skill-path UX fixes; treat as monorepo sync, not a branded semver channel.
  • Windsurf (v3.6.27): Corporate proxy users should upgrade for Windows cert-store fix; local tool symlink write refusal is a concrete foot-gun close.

For teams

  • Enterprise auth & residency: Google Antigravity 1.1.10 Business sign-in (Gemini Enterprise seats, region, admin controls) is the clearest org-readiness signal this window — pilot with a allocated GE seat and confirm admin policy surface before broad rollout.
  • Self-hosted agent runtime: Claude Code self-hosted runners (Team/Enterprise) move session compute onto your machines/containers — define image hardening, network egress, and who may register runners before enabling.
  • Plugin supply chain: Codex portable plugins + marketplaces and Claude archive HTTPS+SHA-256 installs need an allowlist/pinning policy; ban unsigned zip/plugin URLs in shared environments.
  • Approval policy: Codex --approve-for-me plus v0.146.1 safer defaults for cyber-capable models — standardize on the patched CLI, document when auto-review is allowed, and keep human review for high-risk repos.
  • MCP OAuth at work: Cline v4.1.7 pre-registered OAuth clients unblock remote MCP where dynamic registration is disabled — coordinate client IDs with IdP owners.
  • Checkpoint / rewind discipline: Cline desktop transactional workspace restore reduces half-applied agent damage — require checkpoint-backed workflows for shared working trees.
  • Symlink & cert posture: Windsurf’s refuse-write-through-symlink and Windows enterprise CA fixes are relevant to regulated desktops; verify against your TLS-inspecting proxies.

What to ignore

  • Gemini CLI nightlies/previews (v0.56.0-nightly…, v0.54.0-preview.0): caretaker/GCP job churn and duplicate changelog bleed — useful for maintainers, not default install pins.
  • Zed AI v1.15.0-pre: real editor/git UX, but prerelease; wait for stable before fleet policy.
  • xAI Grok 8a14c91 “Synced from monorepo”: commit-level signal without a clean product version — track, don’t standardize on the hash alone.
  • Windsurf HTML changelog: valid security fixes, lower provenance weight than GitHub release tags in this mix.
  • Open proposals at P5 (older Codex 0.144/0.145, Cline 4.0.7, Goose 1.44.0, Zed 1.13.0-pre, etc.): superseded by newer evidence above for day-to-day decisions.

Watch next

  • Codex Plugin marketplace adoption and whether --approve-for-me defaults stay conservative on cyber-capable models after rust-v0.146.1.
  • Claude Code self-hosted runner + archive plugin pinning in real Team/Enterprise rollouts (sandbox/permissions tags were hot in-signal).
  • Cline desktop checkpoint/fork reliability under long-running MCP sessions (SDK abort/queue durability).
  • Antigravity Enterprise admin-control depth once Business sign-in sees production traffic.
  • Goose compaction templates + Gemini model support follow-through.
  • Any stable (non-pre) Zed release shipping git.diff_base.
  • Clearer semver tagging on Grok Build if it remains in the CLI competitive set.

How to use DevAgentRadar

  • Developers: pin versions from the table, apply one concrete flag or workflow change (plugins, checkpoints, self-hosted runner, HTTPS/cert fixes), and discard nightlies unless you are reproducing a listed bug.
  • Teams: turn the same rows into policy — allowlisted plugin sources, approval modes, Enterprise auth, and symlink/proxy hardening — and ignore P5 backlog proposals until they reappear with fresher release evidence.
  • Everything here is judged from the supplied versioned signals only; no invented releases or marketing claims.

Evidence appendix

  • Evidence window: top curated signals (versioned releases preferred)
  • Assistants tracked: 22
  • Signals in this brief: 16
  • Trusted source mix (GitHub/changelog vs HTML): 55%
  • Open improvement proposals: 8
  • Sources reporting errors: 0

High-signal releases used

  • Claude Code v2.1.224: Added self-hosted environments: claude self-hosted-runner turns your own machines or containers into a place Claude Code web, mobile, and desktop sessions can — source

  • OpenAI Codex CLI rust-v0.147.0: Install portable Agent Plugins and search across local, personal, workspace, and remote plugin catalogs. (#36544, #36409, #36919, #36796) · Organize conversatio — source

  • OpenAI Codex CLI rust-v0.146.0: Name new sessions with /new or /clear, pin important threads, and switch between side conversations without closing them. (#34605, #34840, #35011) · Support — source

  • Zed AI v1.15.0-pre: This week's release includes a new git.diff_base setting for choosing whether editor gutters, file-status colors, and git: diff show uncommitted changes aga — source

  • Google Antigravity 1.1.10: Added Business sign-in for Gemini Enterprise accounts, so you can authenticate with a Google Cloud project under Google Cloud terms, use a license seat allocate — source

  • Cline desktop-v0.0.8: Edit any earlier message in a conversation — the app forks the session at that point, rewinds the workspace to that run's checkpoint, and re-runs from your edit — source

  • Cline v4.1.7: Restore the "View Changes" button on completion rows, backed by SDK checkpoints, so you can review everything a task touched from the completion card. · Bring b — source

  • Cline sdk/sdk/v0.0.72 — Prompts queued during a turn now survive being interrupted: they are preserved across user-initiated aborts, drained after a turn aborts itself, and edits made — source

  • OpenAI Codex CLI rust-v0.146.1: Apply safer automatic-review defaults for cyber-capable models and explain permission changes in the terminal interface. (#37057) · #37057 [0.146] Backport safe — source

  • Goose v1.45.0: Stable agent event message identity #10716 · Tab completion and switch provider for model command [#10585](h — source

  • Gemini CLI v0.56.0-nightly.20260808.gcf22ac7e8: Reclassifying Capacity Exhaustion as Terminal Error · feat(caretaker): update Firestore schema with error, and pr_number fields · feat(caretaker-triage): prompt — source

  • OpenCode v1.18.5: Improve Claude adaptive thinking handling across more response shapes. · Avoid OpenAI Responses phase handling that could break some conversations. · Preserve g — source

  • Gemini CLI v0.54.0-preview.0: Changelog for v0.53.0-preview.0 · Changelog for v0.52.0 · chore(release): bump version to 0.54.0-nightly.20260722.gf743ab579 · fix(caretaker): sanitize and wrap — source

  • Gemini CLI v0.54.0: Changelog for v0.53.0-preview.0 · Changelog for v0.52.0 · chore(release): bump version to 0.54.0-nightly.20260722.gf743ab579 · fix(caretaker): sanitize and wrap — source

Open proposals

Was this guidance useful?

Anonymous · one vote per visitor · re-click to clear. Helps improve digests—not a product ranking.