Developer tools

AI SDK added HarnessAgent as a common interface for Pi, Claude, Codex, OpenCode, and other harnesses. Use it to run local or cloud software-factory jobs through official SDKs while subscriptions cover token usage.

MinerU was documented as a local OCR pipeline for PDF, Office, and image-to-Markdown with LaTeX formulas, tables, and 109 languages. The workflow adds mineru -p, mineru-api, Gradio, and an MCP server for Claude Desktop or Cursor.

Mastra published an OpenUI guide, CopilotKit showed Fable 5 agents rendering AG-UI React components, and OpenUI Adam targeted Vercel Eve. The cluster creates a shared component-streaming surface across React, design systems, and filesystem agents.

Grok Build added speech-to-text dictation for coding agents through /voice or Ctrl+Space. Try it to bring Grok-powered real-time voice input into CLI coding workflows.

Browser Use CLI 3.0 shipped direct Chrome DevTools Protocol control through browser-harness with a 6× smaller context path. Try it with Claude Code, Codex, cloud browsers, or local Chrome sessions to cut browser-agent context overhead.

Z.ai released ZCode as its official desktop environment for GLM-5.2, with multi-agent project work, long-running tasks, code review, and clients for macOS, Windows, and Linux. GLM Coding Plan subscribers get a 1.5x quota inside ZCode, while other developers can bring existing subscriptions or API keys.

Firecrawl expanded /monitor from single pages to whole-web tracking, with examples covering filings, competitor changes, hiring, and news alerts. The feature ships across the API, CLI, Playground, and Firecrawl MCP for direct use in agent and search workflows.

Anthropic shipped Claude Code 2.1.198 with Claude in Chrome, background agents that auto-commit and open draft PRs, and a new eval command with ablation and judge-model options. The release also adds AWS upstream failover and retries transient mid-response network drops instead of aborting turns.

Independent toolmakers pushed GLM 5.2 into coding workflows via dcode, Amp plugin modes, and Wafer-backed Next.js routes, while Composio reported it tied or won across 41 real-tool tasks. That matters because GLM is moving from benchmark curiosity into a practical open-weight option for agentic coding and long-running repo work.

Vercel added Dockerfile-based Functions, a Services model for multi-framework apps in one project, and a VCR registry for container images at Ship NYC. The release lets teams deploy OCI images and collocated services with atomic rollbacks, private networking, and active-CPU billing, so Docker-based apps can move without single-runtime constraints.

Anthropic launched Claude Science in beta as a research app with traced artifacts, on-demand environments, and access to more than 60 scientific databases. Modal is already integrated as an elastic compute layer, giving researchers a single workspace for data access, code, and reproducible runs.

Anthropic opened a Claude Desktop beta for Ubuntu and Debian that bundles chat, Claude Code, and Claude Cowork in a native Linux app. It gives Linux users a first-party desktop path into Claude workflows, though Computer Use is still missing from this release.

Vercel raised the maximum package size for Functions on Fluid compute from 250 MB to 5 GB, a 20x increase. The change removes a common deployment blocker for browser automation, larger Python AI stacks, image processing, and heavier backend workloads.

Next.js 16.3 Preview adds major Turbopack gains, including up to 90% less dev-memory use, up to 5.5x faster warm builds, and a Rust React Compiler path that sped route compilation 20-50% in tests. The update matters for longer agent-heavy sessions where dev caches, typecheckers, and coding tools all compete for RAM.

Claude Code 2.1.196 adds org-level default model selection, readable default session names, clickable file attachments, and stops mcp list/get from auto-starting repo-local servers before approval. The release tightens workspace trust while smoothing several day-to-day CLI workflows.

Vercel shipped realtime speech and transcription support in AI Gateway and AI SDK 7, then added Grok voice models through the same interface. The update puts voice agents on the same gateway, WebSocket, and AI SDK stack Vercel already uses for text models.

A day after /goal and thread automations landed in Codex, practitioners started standardizing on /goal specs, /fork or /side detours, and /rewind plus /compact recovery. The pattern matters because verifier design and compaction timing now control how well long runs hold together.

Plannotator v0.21.3 shipped file-scoped comments, a unified review UX, default per-file Ask AI chats, and a more reliable Codex app-server path. It matters because guided reviews and plan checks can now plug into agent workflows with less custom glue.

Microsoft open-sourced SkillOpt, a system that treats agent skill documents as tunable artifacts and improves them against measured task batches. It matters because practitioners are already standardizing shared /research, QA, and packageable skills across harnesses, turning skill files into a new optimization surface alongside models.

Codex users documented thread automations as recurring wake-up calls that preserve thread context, alongside /goal and /btw patterns for steering long-running loops. The workflow matters because teams can schedule check-ins, queue instructions mid-run, and add adversarial review passes without building a separate orchestrator.

OpenAI shipped another Codex desktop update with smoother long-thread scrolling, deeper local history, better settings search, and a hover navigation rail. The release matters because long-running sessions keep your place and copy richer Markdown into Slack.

OpenCode v2 moves its TUI, desktop, and web clients onto a shared backend so sessions stay synced and resource use drops across windows. The beta matters for multi-window agent workflows, though the next build still lacks features.

Datalab’s balanced extraction mode scored 95.9% on a 225-document benchmark and beat Reducto Deep Extract’s 95.1%, according to Vik Paruchuri. The update also adds citations and reasoning, but the benchmark and price comparison are vendor-reported.

Vercel extended the AI SDK Harness API to cover OpenCode and Deep Agents, adding more agent runtimes to the unified interface introduced in AI SDK 7. The change matters because apps can swap supported runtimes without rewriting integration code, though ACP is still awkward for some cloud deployments.

Next.js previewed an agent-focused toolchain with auto-managed AGENTS.md, browser-backed verification, and Skills for cache-component migration and optimization. The release matters because framework guidance, browser introspection, and fix prompts are now packaged directly for coding agents.

Google AI Studio shipped Design Variations, which generates multiple UI directions from an existing build and lets users apply one directly. It matters because builders can branch app presentation without rewriting aesthetic prompts or manually rebuilding layouts.

OpenAI published usage data showing Codex now generates 99.8% of its internal AI output tokens, with sharp growth in legal, support, recruiting, and finance. The report measures agent adoption as delegated parallel work, not just chat inside engineering.

OpenRouter released an MCP server that lets agents query live model pricing, benchmark scores, provider data, docs, and run test inference from the CLI. That replaces stale model knowledge with current routing data inside long-running agent workflows.

v0 Design Systems 2.0 imports components, tokens, providers, and usage patterns from repos, packages, Storybook, Figma, screenshots, and real apps. That lets generated UI target a team's production design system instead of generic components.

Claude Code 2.1.193 routes all shell commands through auto-mode classification, adds live file path autocomplete in bash mode, and can emit assistant-response OpenTelemetry events. It also changes denial logging and response-logging defaults for teams instrumenting the CLI.

Vercel shipped AI SDK 7 with approvals, durability, telemetry, and other production agent primitives. Early adapter feedback points to breaking changes and migration work for SDKs that wrap the old APIs.

Seedance 2.0 rolled out native 4K generation while Seedance 2.0 Mini landed on fal, Replicate, Pika MCP, and ComfyUI. That matters because engineers can now reach the same video model family through APIs, MCP workflows, and local graph tooling instead of a single web surface.

Zed v1.8 added agent.terminal_init_command plus Git, diff, and multi-cursor performance work. The update makes new agent terminal threads easier to bootstrap with project-specific setup and lowers editor overhead.

Genspark turned Build Preview into Genspark Design and merged its AI Designer tooling into one product with Figma uploads, reusable brand systems, and code export. The launch matters because it pushes design-to-code workflows toward editable layered output instead of one-shot mockups.

Baidu released Unlimited OCR as an open-source long-document OCR model with 3B total parameters and 500M active at inference. Early ParseBench testing says it is strong on tables and reading order but weaker on semantic formatting and charts, giving teams a new open-weight OCR option with clear tradeoffs.

Claude Code 2.1.191 introduced /rewind, made stopped background agents stay stopped, and cut streaming CPU use by about 37%. The update changes session recovery and long-running task control, so migrate to the new workflow if you rely on background agents.

OpenRouter released a dedicated Image API that normalizes request shapes across 30-plus models from eight providers. Agents can inspect limits, passthrough options, streaming, and exact per-call cost without hardcoding vendor quirks.

Claude Tag puts Claude into Slack as a teammate that can handle threads, use approved tools, and follow up proactively in selected channels. Team and Enterprise users can try it in beta to keep shared channel context instead of restarting from private chats.

AssemblyAI’s Universal-3.5 Pro Realtime now carries forward the agent side of a conversation to improve live transcription. The release also ships multilingual realtime ASR features, and one early deployment said critical-utterance errors fell from 26% to 9%.

Perceptron’s Files API lets developers upload an image or video once and reference it by ID across later requests instead of resending base64 or URLs. That simplifies repeated multimodal workflows and cuts transfer overhead for video-heavy pipelines.

Claude Code 2.1.186 adds CLI-based MCP auth, automatic assistant replies after ! shell commands, and tighter named-subagent permission checks. The update cuts interactive setup for remote MCP servers and tightens policy-heavy agent workflows.

Vercel rolled out native WebSocket support so Node.js libraries like Socket.IO can run from CDN to Fluid. Existing sessions still reconnect at the 30-minute function limit, so teams should test long-lived connections before migrating.

Claude Design now deploys directly to Vercel with one click. The integration turns design output into a live previewable app without leaving the design flow, extending Claude Design beyond imports and code sync.

Files SDK 2.0 adds a gateway module, browser clients, server adapters, and a shadcn/ui registry around one storage API. The release turns a server-side wrapper into a full-stack file layer with scoped auth, range downloads, and versioned operations.

Simon Willison released the first sqlite-utils 4.0 release candidate with a built-in migrations system and nested transactions. The RC adds minor backward incompatibilities while expanding SQLite workflow automation for scripts and apps.

BrowserCode, Hyper, OpenCode, Together, and other vendors added GLM-5.2 soon after release. That turns the open model into a deployable option across coding, browser automation, and hosted chat.

Hermes now offers a setup path that starts with only a provider, model, file operations, and terminal access. The smaller base gives users a minimal install they can extend manually.

Independent tests put GLM-5.2 near Opus 4.8 and GPT-5.5 on planning and coding, and users shared Claude Code, BrowserCode, dcode, and local-serving recipes. It matters because many engineers are treating it as a daily-driver option for text-heavy coding, though teams still report weaker vision and provider limits.

lift-pdf released an open-source 9B model for schema-constrained document extraction, with code, pip install, playground access, and a 90.2% score on the team's 225-document bench. It matters because the model claims near-Gemini 3.5 Flash accuracy at 9.5s p50, though coverage is still skewed toward Latin-language docs and commercial-use limits remain.

Codex can now hand off an in-progress thread between local and remote machines and bring it back later. It matters because the handoff carries Git history, branches, and uncommitted changes while leaving the destination checkout untouched.