Radar brief — 2026-11-21
Evidence-backed brief across 29 assistants from 40 high-signal releases. Trusted sources ~54%.
Judgment from curated release evidence — not a scrap dump of every vendor page.
What the evidence shows
- New in the last 36 hours: 30 release(s) from 17 tool(s)
- Evidence window: new releases first, then top curated signals (versioned releases preferred)
- Assistants tracked: 29
- Signals in this brief: 40
- Trusted source mix (GitHub/changelog vs HTML): 54%
- Open improvement proposals: 7
- Sources reporting errors: 1
Themes
model× 34cli× 26agent× 22security× 22mcp× 21ide× 15breaking× 14pricing× 12
Notable versioned / high-signal items
-
OpenAI Codex CLI
rust-v0.171.0: Added GPT-6.2 Sol and Astra to model catalogs, with Sol as the default catalog selection. (#62106) · Enabled the TUI workspace by default, with searchable docum — source -
Mistral Vibe
v3.1.4:/mcp <source>has a search field that filters its tools by name and description ·/thinking <level>sets the thinking level directly, without opening the pi — source -
Qwen Code
v0.28.3: See the complete change list below. · No known breaking changes. · feat(web-shell): explain background tasks from the composer chips ([#17398](https://github. — source -
Google Antigravity
1.5.14: Added/thinking(/thinking [on|off], saved as"showThinking"insettings.jsonand off by default) to stream the model's full thinking live in the transc — source -
Claude Code
v2.1.330: Added sign-in to organizations managed in Anthropic's Enterprise Admin Console: afterclaude auth login, Claude Code applies the organization's settings and u — source -
Kilo Code
v7.8.32— (release): #14982fe67f44Than — source -
Cline
sdk/sdk/v0.2.1:read_filesresults are no longer re-cut to the 8k tool-result cap on the way to the provider. · They use the 50k file-content cap, so a read the tool reported — source -
Crush
v0.102.0: There's so much good stuff in this release and we're really proud of all of it. · Read on for more. · At last, you can replace the system prompts with…whateve — source -
Cline
v4.1.27: LM Studio settings have an API key field, for servers with "Require Authentication" turned on. · The MCP tab shows a "read-only" badge on tools whose server dec — source -
Claude Code
v2.1.329: Added a trust dialog warning that says where a folder's hooks are declared · Added a warning in the conversation, in Claude Desktop and other non-interactive ho — source -
Google Antigravity
1.5.13: Added Gemini 4 Neon (Low, Medium, High, and XHigh) to/modeland--modelwhen you sign in with Application Default Credentials (AGY_ADC_AUTH), and added s — source -
Cline
cli-v3.0.75: One-shot runs (a prompt argument or piped stdin) now start in yolo mode instead of act mode, since nobody is at the keyboard to approve tools. · Act mode is kep — source -
Kilo Code
jetbrains/v7.2.1-rc.9: Added native Cloud workspace diffs, so Cloud sessions can show their changes in the IDE diff viewer. · Agent Manager now shows pull request status for Cloud wor — source
For developers
- OpenAI Codex CLI rust-v0.171.0: try the change on one real task before switching defaults.
- Mistral Vibe v3.1.4: try the change on one real task before switching defaults.
- Qwen Code v0.28.3: try the change on one real task before switching defaults.
- Google Antigravity 1.5.14: try the change on one real task before switching defaults.
- Claude Code v2.1.330: try the change on one real task before switching defaults.
- Keep a second assistant ready so vendor downtime is not career downtime.
For teams
- OpenAI Codex CLI: review security/pricing/breaking notes before org-wide enablement.
- Mistral Vibe: review security/pricing/breaking notes before org-wide enablement.
- Qwen Code: review security/pricing/breaking notes before org-wide enablement.
- Google Antigravity: review security/pricing/breaking notes before org-wide enablement.
- Claude Code: review security/pricing/breaking notes before org-wide enablement.
- Treat assistants as a portfolio; standardize only after a pilot.
What to ignore
- Nav/marketing page scrapes without versions
- Duplicate changelog + release entries for the same version
- “Stale” assistants until source health recovers (see Compare)
Watch next
- Open proposals: decide on /proposals
- Filter to My stack so you only see tools you use
- Revisit after the next radar refresh for new high-signal releases
How to use DevAgentRadar
- Read this brief for judgment
- Open linked sources to verify
- Act on proposals (approve / defer / ignore)
Related proposals
- Claude Code v2.1.330 (audience: team, P5)
- Qwen Code v0.28.3 (audience: team, P5)
- Mistral Vibe v3.1.4 (audience: team, P5)
- rust-v0.171.0 0.171.0 (audience: team, P5)
- (release) (audience: team, P4)
- Improve with xAI Grok CLI (Grok Build) c68e39f: Publish harness and TUI open-source (audience: everyone, P2)
- Improve with OpenCode pr-38252-videos: pr-38252-videos PR #38252 verification videos (audience: everyone, P2)
Evidence appendix
- New in the last 36 hours: 30 release(s) from 17 tool(s)
- Evidence window: new releases first, then top curated signals (versioned releases preferred)
- Assistants tracked: 29
- Signals in this brief: 40
- Trusted source mix (GitHub/changelog vs HTML): 54%
- Open improvement proposals: 7
- Sources reporting errors: 1
High-signal releases used
-
OpenAI Codex CLI
rust-v0.171.0: Added GPT-6.2 Sol and Astra to model catalogs, with Sol as the default catalog selection. (#62106) · Enabled the TUI workspace by default, with searchable docum — source -
Mistral Vibe
v3.1.4:/mcp <source>has a search field that filters its tools by name and description ·/thinking <level>sets the thinking level directly, without opening the pi — source -
Qwen Code
v0.28.3: See the complete change list below. · No known breaking changes. · feat(web-shell): explain background tasks from the composer chips ([#17398](https://github. — source -
Google Antigravity
1.5.14: Added/thinking(/thinking [on|off], saved as"showThinking"insettings.jsonand off by default) to stream the model's full thinking live in the transc — source -
Claude Code
v2.1.330: Added sign-in to organizations managed in Anthropic's Enterprise Admin Console: afterclaude auth login, Claude Code applies the organization's settings and u — source -
Kilo Code
v7.8.32— (release): #14982fe67f44Than — source -
Cline
sdk/sdk/v0.2.1:read_filesresults are no longer re-cut to the 8k tool-result cap on the way to the provider. · They use the 50k file-content cap, so a read the tool reported — source -
Crush
v0.102.0: There's so much good stuff in this release and we're really proud of all of it. · Read on for more. · At last, you can replace the system prompts with…whateve — source -
Cline
v4.1.27: LM Studio settings have an API key field, for servers with "Require Authentication" turned on. · The MCP tab shows a "read-only" badge on tools whose server dec — source -
Claude Code
v2.1.329: Added a trust dialog warning that says where a folder's hooks are declared · Added a warning in the conversation, in Claude Desktop and other non-interactive ho — source -
Google Antigravity
1.5.13: Added Gemini 4 Neon (Low, Medium, High, and XHigh) to/modeland--modelwhen you sign in with Application Default Credentials (AGY_ADC_AUTH), and added s — source -
Cline
cli-v3.0.75: One-shot runs (a prompt argument or piped stdin) now start in yolo mode instead of act mode, since nobody is at the keyboard to approve tools. · Act mode is kep — source -
Kilo Code
jetbrains/v7.2.1-rc.9: Added native Cloud workspace diffs, so Cloud sessions can show their changes in the IDE diff viewer. · Agent Manager now shows pull request status for Cloud wor — source
Open proposals
- Claude Code v2.1.330 (audience: team, P5)
- Qwen Code v0.28.3 (audience: team, P5)
- Mistral Vibe v3.1.4 (audience: team, P5)
- rust-v0.171.0 0.171.0 (audience: team, P5)
- (release) (audience: team, P4)
- Improve with xAI Grok CLI (Grok Build) c68e39f: Publish harness and TUI open-source (audience: everyone, P2)
- Improve with OpenCode pr-38252-videos: pr-38252-videos PR #38252 verification videos (audience: everyone, P2)
Anonymous · one vote per visitor · re-click to clear. Helps improve digests—not a product ranking.