The Oct 21 release fixes sessions with many MCP tools showing a real API error as "prompt is too long" and running an unneeded compaction; per TurboAI, a 400 or 422 that names deferred tools now retries without them. It also fixes claude -p inside cloud sessions reporting "Not logged in" when started without the session's environment variables.
Launched Oct 20 for Pro, Max, and Enterprise subscribers, the new mode runs planning, reasoning, and file work on device with Qwen 3.6 or the post trained PPLX Pro, working across local files plus Notes, Messages, Mail, and Numbers when you grant access. Perplexity recommends 36GB or more of unified memory, and you can still authorize cloud escalation for browsing, fresh info, or connected apps (9to5Mac)…
The Oct 20 release gives dynamic workflow subagents native structured output on supported models, adds an assistedApprovalInstructions setting to steer Assisted Permissions, and stops local or loopback providers in providers.json from overriding Auto or your saved model. It also makes headless and SDK sessions skip MCP servers whose saved tokens need interactive sign in, and fixes the first prompt hanging silently when no model is selected…
As of Oct 19, assigning a Linear issue to Amp (or mentioning @Amp in a comment) starts a Linear Agent Session backed by an orb that runs as you, with progress and replies in Linear and the same thread viewable in Amp. A workspace admin installs Amp in Linear and can route Linear teams or projects to specific Amp projects, and each user connects Linear under Settings → Integrations & MCP…
Announced Oct 20, admins opt in from the Cloud Console and usage is billed pay as you go outside the pooled Gemini Enterprise quota, with both models at four thinking levels across Antigravity 2.0, the CLI ( agy model claude opus 5 5 medium ), and the VS Code, JetBrains, and Visual Studio extensions. Background Gemini models still handle context management and summaries, so those tokens can add charges too (docs)…
The MIT licensed tool (GitHub) records prompts, tool calls, and edits from 20+ coding agents in one local first history, then turns sessions you review into skills that any later agent can pull in through MCP or Agent Skills. Asymptote Labs, the company behind it, announced a $5.5M seed led by Tower Research Ventures on Oct 20, and its paid platform adds blocking of risky agent actions (The Next Web)…
Spotted something we missed? Start a thread.
