Radar brief — 2026-08-14
Agent runtimes tighten control surfaces: plan-mode hard blocks, safer auto-approvals, self-hosted runners, and portable plugins dominate this window—while MCP auth and enterprise sign-in keep expanding the team footprint.
What the evidence shows
Versioned GitHub releases carry most of the weight here (Codex, Claude Code, Cline, Goose, Zed, Gemini CLI). One HTML changelog (Windsurf) and several prereleases/nightlies are noisier and should be weighted lower.
git.diff_base chooses gutter/diff baseline (HEAD vs default-branch merge base); macOS/Linux Wayland drag-out from Project Panel; linked editing + Emmet in JSX/TSX bodiesclaude self-hosted-runner for Team/Enterprise web·mobile·desktop sessions on your machines; archive plugin source (HTTPS zip + optional SHA-256 pin); ANTHROPIC_BEDROCK_REGION_PREFIX for Bedrock inference profiles--approve-for-me auto-reviewed approvals; import/sync Cursor & Claude skills/conversations without dupesrun.failedPrerelease / thin signals (context only): Gemini CLI v0.56.0-preview.1 and a v0.56.0-nightly.* build (caretaker/triage machinery); Codex rust-v0.148.0-alpha.17 (tag only); xAI Grok CLI commit be71313 (monorepo sync—presence protocol, worktree/fetch tweaks).
For developers
- Codex CLI
rust-v0.147.0: Try portable Agent Plugins and skill import from Cursor/Claude if you juggle multiple assistants—verify catalogs (local/personal/workspace/remote) before relying on them in a clean clone. Treat--approve-for-meas convenience with review, not a blank check; pair with therust-v0.146.1safer cyber-model defaults. - Cline
v4.1.4: Relearn plan mode—file mutation, in-place editors, redirects, mutating git, and package installs are hard-refused. Drive plan→act yourself; use skills in the slash menu and resolve same-name command clashes explicitly. - Cline Desktop
desktop-v0.0.8: Use mid-thread message edit + checkpoint rewind for “branch the run” workflows; lean on transactional restore when experimenting so a failed rewind does not half-apply. - Goose
v1.46.0: Turn on per-message usage/cost/TTFT stats to catch expensive loops early; streaming shell helps debug long commands without waiting for full completion. - Zed AI
v1.15.0: Setgit.diff_basedeliberately (uncommitted vs HEAD vs branch-vs-default merge base) so agent-driven diffs match how you review PRs. - Claude Code
v2.1.224: If you package internal plugins, preferarchive+ SHA-256 pin over ad-hoc git/npm. Use cancel-and-confirm when paste removal rewrites a command. - Gemini CLI
v0.55.1: Pull for the sensitive-path blocklist and at-reference path fixes before the next large repo walk.
For teams
- Self-hosted / enterprise path: Claude Code self-hosted runners (Team/Enterprise) and Antigravity 1.1.10 Business/Gemini Enterprise sign-in are the clearest org-control moves—map them to existing seat, region, and admin-policy requirements before pilot expansion.
- Approval & plan policy: Codex’s safer cyber-model auto-review defaults + Cline’s hard plan-mode shell blocks are concrete policy levers. Document which flag/mode is mandatory on shared agents; do not rely on model “good behavior” for write isolation.
- MCP authenticity: Windsurf’s fixes for self-hosted GitLab and older Atlassian MCP instances are a reminder to re-test SSO/self-hosted MCP after client bumps—especially where HTML changelogs are the only note.
- Plugin supply chain: Claude
archiveHTTPS zip + optional SHA-256 pinning and Codex portable plugins both need an internal allowlist and hash review path before wide rollout. - SDK/host stability (Cline
sdk/v0.0.72): If you embed Cline, validate abort/queue/run.failedbehavior across hub restarts so orchestrations do not silently drop work. - Open proposals (all P5): Several improve/review items lag current tags (e.g. Codex 0.144/0.145, Cline v4.0.7, Goose v1.44.0, OpenCode v1.18.5, Zed pre). Refresh proposals against the versions in this brief before assigning owners.
What to ignore
- Codex
rust-v0.148.0-alpha.17: Tag-only alpha—no actionable changelog body in evidence. - Gemini
v0.56.0-preview.1/ nightlygcf22ac7e8: Prerelease caretaker/triage/Cloud Run work; not a stable adoption signal. - xAI Grok CLI
be71313: Monorepo sync commit without a clean semver release narrative; treat as telemetry, not a pin target. - OpenCode
v1.18.10UI polish (toasts, tabs, session button): fine to take opportunistically, not fleet-priority versus security/plan-mode/plugin changes elsewhere. - Marketing framing around “+N more changes” without named behaviors—stick to the concrete bullets above.
Watch next
- Whether Codex
--approve-for-me+ cyber-model review defaults stay restrictive after the0.148alpha line lands a full release notes body. - Cline plan-mode hard blocks: any escape hatches or false positives on legitimate read-only investigation tooling.
- Claude Code self-hosted runner adoption limits (plans, sandbox/permissions surfaces already tagged in-signal) and real-world
archivepin workflows. - Gemini CLI path-blocklist effectiveness on case-variant sensitive paths post-
v0.55.1, and what actually ships from the 0.56 caretaker previews. - Goose cache-safe turn assembly / unrolled loop: impact on long MCP+shell sessions and cost visibility accuracy.
- Windsurf MCP auth: confirm the GitLab/Atlassian fixes on a GitHub-comparable source when possible (current note is HTML changelog only).
How to use DevAgentRadar
- Read What the evidence shows as the only source of truth for this brief—versions and behaviors are drawn from the curated release signals, not vendor blogs.
- Developers: pick one assistant you already run and apply a single concrete action (flag, mode, or pin) from that section today.
- Teams: convert approval, plan-mode, self-hosted, and plugin-pin items into policy checks; refresh open P5 proposals against these newer tags before prioritizing work.
- Discount prereleases, nightlies, bare alphas, and HTML-only notes until they reappear as stable, versioned GitHub-style evidence.
- Re-scan next window for follow-through on Codex 0.148, Gemini 0.56 stable, and any hardening after the plan-mode / auto-approve shifts.
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
-
Zed AI
v1.15.0: This week's release includes a newgit.diff_basesetting for choosing whether editor gutters, file-status colors, andgit: diffshow uncommitted changes aga — source -
Claude Code
v2.1.224: Added self-hosted environments:claude self-hosted-runnerturns 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 -
Goose
v1.46.0: Unrolled agent loop #9574 · Cache-safe request assembly with append-only turn context and declared cache semant — 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 -
Gemini CLI
v0.55.1: fix/verify release npm ci ignore scripts · fix(ci): prevent workspace binary shadowing in release verification · Feat/tool registry discovery · fix(ci): prevent — 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
v4.1.4: Recognize Chutes as a provider. · Show skills alongside workflows in the slash command menu, and disambiguate commands that share a name instead of letting one — source -
Cline
sdk/sdk/v0.0.72— 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 -
Gemini CLI
v0.56.0-preview.1: Changelog for v0.55.0-preview.1 · chore(release): bump version to 0.56.0-nightly.20260806.g761f604c1 · Changelog for v0.54.0 · Reclassifying Capacity Exhaustion — 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 -
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 -
xAI Grok CLI (Grok Build)
be71313— Synced from monorepo: Synced from monorepo · Presence protocol end-to-end: live presence updates through the gateway to client presence tiers · Worktree: narrow standalone origin fet — source -
OpenCode
v1.18.10: Discover available Modal models automatically. (@devennavani) · Prevent adding the same attachment more than once. · Always show the new session button. · Impro — source
Open proposals
-
Improve with OpenAI Codex CLI rust-v0.145.0: rust-v0.145.0 0.145.0 (audience: team, P5)
-
Improve with Cline v4.0.7: v4.0.7 (audience: team, P5)
-
Improve with xAI Grok CLI (Grok Build) 8adf901: Synced from monorepo (audience: team, P5)
-
Improve with OpenCode pr-38252-videos: pr-38252-videos PR #38252 verification videos (audience: everyone, P5)
-
Improve with OpenAI Codex CLI rust-v0.144.6: rust-v0.144.6 0.144.6 (audience: developer, P5)
-
Review breaking changes in Zed AI v1.13.0-pre (audience: team, P5)
-
Improve with OpenCode v1.18.5: v1.18.5 (audience: team, P5)
-
Improve with Goose v1.44.0: v1.44.0 (audience: team, P5)
Anonymous · one vote per visitor · re-click to clear. Helps improve digests—not a product ranking.