Vibecodingwiki
Verified launches, funding, acquisitions, tools, security updates, and community stories from the VibeCoding ecosystem.
A source-linked record of notable product releases, funding, acquisitions, launches, security developments, and community projects. Stories are checked daily; updates are published only when the evidence is specific and dated.
Last verified 15 October 2026
Epherem launched a free scanner for AI-built applications that checks authentication, database access, exposed API keys, and vulnerable packages from ZIP uploads or connected GitHub repositories. It provides plain-language findings and repair prompts for tools including Claude Code, Lovable, and Cursor; Product Hunt ranked it #19 with 77 points.
Sources: Product Hunt, Epherem
Reason launched a coding agent usable from its Mac app, terminal, web interface, or Slack that can write code, run tests, and open pull requests. The company reports a 65.1 score on its published GPT-6 Astra Coding Agent Index evaluation versus 62.5 for Codex, with 22% lower cost per solved task; Product Hunt ranked it #8 with 111 points.
Sources: Product Hunt, Reason
Tiden launched a verification layer that captures requirements before an agent edits code, versions those requirements like code, runs environment-based tests, and connects releases back to the originating intent. Product Hunt listed it at #5 for the day with 192 points.
Sources: Product Hunt, Tiden
GitHub introduced workflow lockfiles that pin Actions dependencies and reusable workflows to reviewed commits, reducing the risk that moved tags or branches introduce unexpected code. Administrators can also require lockfiles at enterprise, organization, or repository scope.
Sources: GitHub Changelog
GitHub updated its multi-model HydraFusion orchestration workflows, added HydraFusion Max for more demanding tasks, and made both options available in VS Code, the GitHub Copilot app, and Copilot CLI in research preview.
Sources: GitHub Changelog
Codex Remote launched a Mac menu-bar app for provisioning Claude Code and Codex on user-controlled cloud machines. Product Hunt describes support for multiple cloud providers, session monitoring, project sync, and opt-in MCP access, and listed the launch at number 30 for the day.
Sources: Product Hunt
Codemind launched a Product Hunt service that accepts coding tasks from agents such as Claude Code, Codex, Cursor, and OpenCode, then returns tested files against supplied criteria. Product Hunt listed it at number 19 for the day.
Sources: Product Hunt
OpenCode launched version 2.0 on Product Hunt, positioning the product as an open-source coding agent available through terminal, desktop, and web interfaces. Product Hunt listed OpenCode 2.0 at number two for the day on October 13, 2026.
Sources: Product Hunt, OpenCode GitHub
GitHub added startup-time refresh enforcement for managed Copilot policies and minimum client-version requirements beginning with sandbox settings. Clients that cannot confirm current policy can fail closed instead of running with stale settings.
Sources: GitHub Changelog
GitHub added Google’s Gemini 4 Argon to Copilot for complex coding, agentic, and reasoning tasks. Availability includes Copilot CLI, coding agent, the Copilot app, and major IDE integrations, with gradual rollout and usage-based billing.
Sources: GitHub Changelog
The open-source DeepCode agent shipped v2.3.1 with streaming Markdown answers in its terminal UI, a persistent Windows background service, improved session compaction, and fixes for destructive-command detection and private-address access. The release also moved pypdf, PyJWT, urllib3, and multidict to patched versions.
Sources: DeepCode GitHub repository
Pi Agent Desktop 0.10.0 upgraded its Pi SDK dependencies to 1.1.0, adding project-level MCP overrides, cancellable MCP OAuth requests, tool-duration reporting, and Claude Haiku 5.5 support. The release also pinned a patched brace-expansion version, enforced HTTPS for global provider credentials except loopback, and refreshed several security-related dependencies.
Sources: GitHub Releases
Dropstone launched Dropstone CLI on Product Hunt with persistent project context, cross-file editing, command and test execution, learned instructions, subagents, MCP tools, and GitHub issue-to-pull-request workflows. Product Hunt recorded 80 upvotes and a #10 daily ranking on October 12, 2026.
Sources: Product Hunt
Bevel launched Hexis Cloud, a hosted workspace for shared company context, reusable skills, MCP tool access, permissions, and agent identity. Product Hunt recorded the launch on October 12, 2026 and ranked it #2 for the day. The launch is company-reported and platform-ranked.
Sources: Product Hunt
Hypervibe launched a native Mac workspace that places Claude Code, Codex, Gemini, and other CLI agents side by side on an infinite board. The product includes agent status, live previews, notes, teams, voice control, and Git project context. Product Hunt recorded Hypervibe at #18 for its launch day.
Sources: Product Hunt
AgentDock launched a macOS utility that aggregates questions and permission prompts from Claude Code, Codex, and other terminal agents. It supports multiple terminal environments, local session recaps, and optional iPhone access over Wi-Fi. Product Hunt recorded AgentDock at #33 for its launch day.
Sources: Product Hunt
Farol launched a free, MIT-licensed macOS terminal for running Claude Code and Codex in parallel. It provides agent status indicators, Git worktrees, diff review, conflict resolution, and pull-request workflows. Product Hunt recorded Farol at #23 for October 11, 2026.
Sources: Product Hunt, Product Hunt daily leaderboard
Buda launched API Claws, a developer-facing API for hosted agents with persistent Drive memory, model routing, managed runtimes, sessions, observability, and embeddable experiences. Product Hunt recorded the launch at #13 for October 10, with 78 points.
Sources: Product Hunt, Buda
GitGlow launched on Product Hunt with a local Git workspace focused on staging, review marks, verification checks, concurrent workspaces, and managed agent workflows. Product Hunt recorded it at #12 for October 10, with 78 points.
Sources: Product Hunt, GitGlow
AgentPlus released v0.4.1, an open-source desktop utility for configuring providers across 18 AI coding agents, including Codex, Claude Code, OpenCode, Gemini CLI, and Qwen Code. The release adds support for Codex 26.1007’s Fast interface behavior.
Sources: AgentPlus
Qoder replaced Desktop’s Performance mode with Fast on October 10. The company says Fast preserves output quality while reducing response time from about 8 seconds to 3 seconds and credit consumption from 1.1x to 0.8x.
Sources: Qoder product announcement
The open-source Coucou project released Linux 0.3.0 with Wayland global shortcuts, terminal reopening, hover-to-open behavior, and additional desktop customization. Coucou monitors multiple coding agents and supports approvals and session controls across desktop and iPhone surfaces.
Sources: Coucou on GitHub
GitHub’s October 5 weekly release adds side-by-side views for agent sessions in VS Code and a chat command for reviewing and deleting inactive session worktrees. The release also documents separate Copilot-license and repository accounts in the Copilot app.
Sources: GitHub Changelog
AgentGuard launched a free scanner for SKILL.md files that reports capabilities, security findings, quality signals, and a verdict before installation. The maker states that the analysis is limited to the skill document and does not inspect an agent's runtime tools or permissions.
Sources: Product Hunt, AgentGuard
BotBus launched a phone and Apple Watch interface for managing local Codex, Claude Code, Hermes, OpenCode, DeepSeek Harness, and other agents across computers. Its Product Hunt listing says it supports task continuation, approvals, file and terminal access, and connectors for macOS, Linux, and Windows.
Sources: Product Hunt, BotBus
pmtui launched a terminal workflow that watches multiple agent sessions, rebuilds continuation prompts from each agent's goal and status, and can use a second model to handle routine decisions. Product Hunt listed it at #45 for the day.
Sources: Product Hunt, pmtui GitHub repository
Hallmonitor launched a free, open-source Mac and terminal utility that monitors Claude Code and Codex sessions across local and remote machines, with notifications and controls from the notch or menu bar. Product Hunt listed it at #33 for the day.
Sources: Product Hunt, Hallmonitor GitHub repository
GitHub added controls letting organizations bill Copilot code reviews to an organization cost center instead of member quotas and restrict review requests to users with organization- or enterprise-provided Copilot licenses.
Sources: GitHub Changelog
OpenSwarm launched a desktop app for macOS and Windows that coordinates multiple agents across app-building, browsing, research, and other tasks. Product Hunt recorded it as the #1 product of the day on October 8, 2026.
Sources: Product Hunt, Product Hunt awards
Copilot CLI version 1.0.94-0 can discover supported models from a running local Ollama instance in the existing model picker. Users must install Ollama and the model separately, and selected models must support tool calling and streaming.
Sources: GitHub Changelog
GitHub made Claude Haiku 5.5 generally available across Copilot surfaces including IDEs, CLI, cloud agent, app, web, and mobile. GitHub positions it for fast, high-volume work such as subagents, quick edits, and terminal tasks.
Sources: GitHub Changelog
GitHub introduced a fine-tuned model for context-aware secret detection. Existing AI-detected secret alerts switch to the model at no additional charge, while AI checks for push protection and Copilot security reviews are entering preview with AI Credit usage.
Sources: GitHub Changelog
GitHub made local sandboxing generally available in Copilot CLI, the Copilot app, and VS Code Agent Host sessions. Policies can restrict filesystem, network, credential, and tool access across Windows, macOS, and Linux.
Sources: GitHub Changelog
GitHub disclosed that IDEs moving Copilot agent sessions to the Copilot SDK caused agent activity to be omitted or misattributed in usage reports. A fix is available in Visual Studio Code 1.139.0 and later, with updates for other IDEs planned through November 2026; GitHub said missing historical data cannot be backfilled.
Sources: GitHub Changelog
GitHub added AI Scan for pull requests enablement status to the security overview for organization and enterprise administrators. The view now reports enabled and not-enabled repository counts, supports filters, and includes AI Scan status in coverage CSV exports.
Sources: GitHub Changelog
GitHub made stacked pull requests generally available across github.com plans, adding stack-aware rebasing, signed replacement commits, merge-queue support, navigation, webhooks, and expanding agent and CLI workflows. GitHub reported that repositories using stacks saw 9% more merged code than peers during preview, while top repositories saw a 5% improvement in time to merge.
Sources: GitHub Changelog
OpenBot launched on Product Hunt as a cross-platform workspace for connecting Claude, ChatGPT, Grok, OpenCode, and custom agents, with role assignment, delegation, and live team collaboration. Product Hunt ranked it second for the October 6 daily launch leaderboard with 330 points.
Sources: Product Hunt
Rill Browser launched on Product Hunt as a Mac browser that lets users turn web pages into tasks for Claude Code or Codex while retaining their existing subscriptions. Product Hunt ranked it first for the October 6 daily launch leaderboard with 403 points.
Sources: Product Hunt
Reviu launched a rebuilt desktop workspace centered on reviewing code produced by Claude Code, Codex, Gemini, and other ACP agents. The product supports durable sessions, per-turn diffs, inline comments sent back to agents, checkpoints, undo, worktrees, and GitHub pull-request workflows.
Sources: Product Hunt
devpit launched a local-first desktop app for running and monitoring Claude Code agents. Its Product Hunt listing highlights a visual board, per-call cost tracking, multi-project orchestration, and an open-source distribution model.
Sources: Product Hunt
Opengeni launched a self-hostable agent infrastructure platform with durable sessions, replayable histories, managed sandboxes, governed credentials, human approvals, integrations, and cost visibility. The Product Hunt launch page lists the project as open source and describes a managed cloud priced at model cost plus 5 percent, with the first 100 users offered 100 dollars in cloud credit.
Sources: Product Hunt, GitHub
Anthropic shipped Claude Code 2.1.290 with richer plugin-hook events, managed-agent onboarding commands, session attach and log commands, and numerous reliability and security fixes. The release also hardens symlink handling, permission checks, MCP controls, sandbox behavior, and restricted sessions.
Sources: Anthropic Claude Code
GitHub released ReviewBench, an open benchmark built from 219 pull requests across 19 languages and modeled on 103.9 million GitHub pull requests. The benchmark includes multi-source findings, severity labels, precision and recall metrics, a public dataset, and a leaderboard for comparing code-review agents.
Sources: GitHub Blog, ReviewBench
Aperture launched on Product Hunt with staged diffs and pre-apply checks for parsing, imports, types, browser preview rendering, and tests. The maker says failed checks remain visible and changes are not applied automatically; Product Hunt listed it at #32 for the day with 61 points.
Sources: Product Hunt, Aperture repository
Sente launched on Product Hunt as a terminal and phone coding agent with two-way voice interaction, access to 404 models through one account, and an MIT-licensed open-source implementation. Product Hunt listed it at #14 for the day with 81 points.
Sources: Product Hunt, Sente repository
The community-maintained Agentic Awesome Skills project released version 18.14.0 with two reviewed skills for container deployment and end-to-end browser-extension delivery, replaced a vulnerable web-app build dependency, and reported a catalog size of 2,640 skills for Claude Code, Cursor, Codex CLI, and Gemini CLI.
Sources: Agentic Awesome Skills GitHub release, Agentic Awesome Skills repository
Tracer Cloud released OpenSRE 0.1.2026.10.4, an open-source AI SRE framework for investigating production incidents across cloud and Kubernetes tooling. The release adds lone-plan execution and improves Codex onboarding with isolated agent state, warmer first turns, and effort configuration.
Sources: OpenSRE GitHub release, OpenSRE repository
Cubicle launched an open-source observability tool that presents Claude Code, Codex, Gemini CLI, and Paperclip agents in a live office view. It includes agent status, approval prompts, task KPIs, Telegram alerts, phone access, and a read-only design intended to prevent the monitor from changing agent systems.
Sources: Product Hunt, Cubicle GitHub repository
Singularity launched a desktop workspace that turns coding tasks into isolated tickets and Git worktrees, allowing agents to run across repositories in parallel before developers review and merge the diffs. Its maker reports 227 tickets shipped across 11 projects, with up to eight agents running at once during internal use.
Sources: Product Hunt, Meteor Factory
CodeGraph 1.6.2 improved symbol resolution across major languages, added route and framework coverage for tools including Angular, Rails, Flask, Expo, Nuxt, and Astro, and improved MCP reconnection and large-project indexing for coding-agent workflows.
Sources: CodeGraph GitHub releases
The open-source nanoMuse project released versions 0.1.28 through 0.1.31, adding desktop Harness installers and a Muse-style desktop across Windows, macOS, and Linux. Version 0.1.31 documents 27 audit fixes, including first-frame WebSocket authentication, signed file links, approval controls, and scrubbed environments for coding CLIs.
Sources: nanoMuse GitHub repository
ZooWork launched a builder and managed API for domain experts and developers to create, deploy, and deliver AI agents. Product Hunt ranked it first for the day on October 3, 2026.
Sources: Product Hunt, ZooWork
Maintainers can now post comments on repository security advisories that are visible only to users with write access. GitHub records views in the audit log and makes confidential comments available through GraphQL, but not the REST API.
Sources: GitHub Changelog
GitHub added public-preview REST endpoints to list, read, create, and edit comments on repository security advisories, including advisories created from private vulnerability reports. The API also exposes comment counts for advisory records.
Sources: GitHub Changelog
GitHub completed its staged rollout of stateless installation tokens. Newly minted tokens now use the longer ghs_APPID_JWT format, while permissions, repository scope, expiration, and the token endpoint remain unchanged.
Sources: GitHub Changelog
GitHub deprecated Gemini 3.5 Flash, Gemini 3.6 Flash, Kimi K2.7 Code, and Claude Opus 4.7 across Copilot Chat, inline edits, ask and agent modes, and code completions. GitHub recommends newer replacement models for each.
Sources: GitHub Changelog
GitHub now supports requesting Copilot code reviews through REST and GraphQL APIs, with an optional review-effort setting. Balanced is now the default effort level for Copilot code review.
Sources: GitHub Changelog
GitHub made its asynchronous merge API generally available for individual and stacked pull requests, merge queues, and direct merges. The API processes merges asynchronously and is now the recommended programmatic merge path.
Sources: GitHub Changelog
GitHub’s September VS Code release summary adds scheduled agent automations, agent merge, pull-request creation from agent sessions, Dev Container sessions, improved session organization, and preview access to HydraFusion model and workflow selection.
Sources: GitHub Changelog
GitHub added structured private vulnerability-report forms requiring summary, details, proof of concept, and impact by default. Repository owners can customize forms, require CWE assignments, and let reporters disclose AI assistance.
Sources: GitHub Changelog
GitHub released public-preview dynamic workflows for Copilot CLI, the GitHub Copilot app, and the Copilot SDK. Developers can define multi-step, sequential, parallel, checkpointed, and reusable agent orchestrations in code.
Sources: GitHub Changelog
GitHub made computer use available in public preview in Copilot CLI and the Copilot app for macOS and Windows. Copilot can read app context, click controls, enter text, navigate workflows, and operate legacy or GUI-only software with user approval and organization controls.
Sources: GitHub Changelog
Styx launched on Product Hunt as a Mac agentic development environment that brings repositories and coding agents into one window. Its maker says the product uses branch isolation, approval prompts, Touch ID for live actions, ephemeral credentials, and an immutable activity log to control agent access to production systems.
Sources: Product Hunt
Campfire launched on Product Hunt as an open-source, harness-independent workspace for consolidating agent sessions, context, decisions, and progress. The maker positions it as a coordination layer for users running Claude Code, Codex, OpenCode, and other coding agents in parallel.
Sources: Product Hunt
Rinkata launched on Product Hunt with a workspace connecting goals, specifications, decisions, and delivery evidence for humans and AI coding agents. Its maker says agents can read and write this context through MCP and reports that the product was built with more than 2,000 tickets, 400 specifications, and 200 decisions.
Sources: Product Hunt
Lovable says its security team found a TanStack Start vulnerability that could let a crafted link execute unwanted JavaScript in affected apps. The company reported the issue to maintainers, added firewall mitigations, and is automatically rolling patched dependencies into affected projects.
Sources: Lovable
iFixAi launched an agent-auditing service with 250 inspections across 69 misalignment categories and GitHub or MCP connectivity. Product Hunt recorded it as the number-one daily launch with 390 points; the maker also reported 15,000-plus open-source stars before the launch.
Sources: iFixAi Product Hunt launch, Product Hunt daily leaderboard, iFixAi website
AWS and OpenAI introduced Bedrock Managed Agents in preview, providing durable sessions, code execution, MCP server connections, reusable skills, IAM identities, human approval for consequential actions, and CloudTrail activity records.
Sources: AWS What's New
Google released Gemini CLI 0.62.0 with structured MCP tool-call formatting, Gemini 3.8 Flash and 3.5 Flash-Lite support, OAuth refresh-token fixes, PTY lifecycle improvements, and broader terminal reliability changes.
Sources: Google Gemini CLI release
Codex 0.159.0 introduced opt-in instant_interrupt input steering, improved session and transcript workflows, native Mermaid rendering, and protections for .aws directories under writable roots.
Sources: OpenAI Codex release
Anthropic shipped Claude Code 2.1.285 with a command to configure plugin options, install-time MCP server settings, the managed allowedProviders policy, a switch to disable WebFetch, and fixes for sandbox approvals and plugin isolation.
Sources: Anthropic Claude Code release
CybeDefend launched VibeDefend on Product Hunt, offering agent-integrated security rules, secret protections, diff scanning, and command guards for tools including Claude Code, Cursor, Windsurf, Copilot, and Codex. Product Hunt records the launch at #5 of the day.
Sources: Product Hunt, CybeDefend
GitHub added Anthropic’s Claude Sonnet 5.5 to Copilot for Pro, Pro+, Max, Business, and Enterprise users, including the coding agent, Copilot CLI, major IDEs, github.com, and mobile. GitHub says rollout is gradual and usage-based billing follows provider list pricing.
Sources: GitHub Changelog
C1.ai launched AppHub, a self-hosted platform that deploys, authenticates, and manages permissions for internally built applications. Its public repository and skills file allow an AI coding agent to build and publish a governed app from a prompt, while applications appear in C1.ai’s identity graph.
Sources: C1.ai, C1 announcement via GlobeNewswire, C1 AppHub repository
Base44 launched Base Code, a browser-based shared cloud workspace that connects GitHub repositories to isolated branches, live previews, AI-assisted changes, and standard pull requests. The product is available to all builders.
Sources: Base44, Base44 announcement via GlobeNewswire
The open-source Emdash agent workspace released v1.2.7 with saved model and permission settings across tasks, direct switching between its chat and terminal interfaces, improved update downloads, more reliable project opening, and better recovery for interrupted sessions.
Sources: Emdash GitHub release v1.2.7, Emdash GitHub repository
MakerMap launched a map-based community directory that lets people discover makers and products by location, niche, interests, and activity. The creator reports more than 4,000 makers and products; Product Hunt ranked MakerMap fourth on September 26, 2026, and its build record credits Claude Code as a co-builder.
Sources: Product Hunt launch page, Product Hunt products-used page, Product Hunt launch awards
Chit launched a 777 KB macOS app and CLI that reads local Claude Code transcripts, groups completed work by project, and formats it for standups or timesheets. The maker reports that the binary has no network code, account requirement, API key, or telemetry; Product Hunt ranked it third on September 26, 2026.
Sources: Product Hunt launch page, Product Hunt September 26 leaderboard
Hemory launched on Product Hunt with an app that continuously organizes conversations into searchable memory and exposes that context to Claude, Codex, Cursor, and other agents through MCP. Product Hunt ranked it first on its September 26, 2026 leaderboard.
Sources: Product Hunt launch page, Product Hunt launch awards
CodeQL 2.27.1 adds new C and C++ security queries, supports Kotlin 2.4.20, improves Go, JavaScript, Rust, and C# analysis, and updates several data-flow and framework models. GitHub says the release is automatically deployed to code-scanning users on github.com.
Sources: GitHub Changelog, CodeQL Changelog
GitHub’s enterprise and organization Copilot usage reports now break pull-request review time into ready-to-first-review, first-to-final-review, and final-review-to-merge stages, with median and 90th-percentile measurements. The new data is available in repos-1-day reports and builds forward from the release date without backfill.
Sources: GitHub Changelog
GitHub Copilot in Slack and Microsoft Teams can now use richer conversation context, including supported files, attachments, images, forwarded messages, and thread history. The update also adds model switching, duplicate-issue checks, source-conversation links, and improved handling of long-running tasks.
Sources: GitHub Changelog
GitHub’s agentic autofix can now use enabled Copilot Memories as repository-specific context when resolving security alerts and store successful fix patterns for future alerts. Both agentic autofix and Copilot Memory remain in public preview.
Sources: GitHub Changelog
Microsoft announced a redesigned Copilot organized around Home, Code, and Autopilot. Code enables natural-language creation of applications and automations, while Microsoft is introducing Copilot Managed Runtime for governed hosting. Home and Code will roll out gradually through the Microsoft Frontier program, with Autopilot expanding to private preview later this month.
Sources: Microsoft News Center, Microsoft Copilot Blog
CtrlOps 1.0 launched with server hardening scores, PDF audit reports, access review and revocation, role management, searchable logs, and GitHub-based deployment. Product Hunt ranked the launch third on September 24, 2026.
Sources: Product Hunt
GitHub Security Lab published an open-source taskflow that uses an LLM agent to identify entry points, create harnesses, run AFL++, improve coverage, triage crashes, and generate vulnerability reports for C and C++ repositories. GitHub warns that it executes arbitrary build commands and should run only in a disposable environment.
GitHub added a public preview that can require interactive reauthentication or a multifactor challenge before managed enterprise users perform sensitive actions such as creating tokens, editing webhooks, changing security settings, or viewing recovery codes.
Sources: GitHub Changelog
GitHub introduced a global policy for generally available Copilot features and supported client capabilities. Administrators can choose whether eligible features are enabled, disabled, or left to organizations, with the policy taking effect on October 22, 2026.
Sources: GitHub Changelog
Floot launched an MCP connector that provides Claude and ChatGPT with a hosted workspace for building full-stack apps, including databases, user accounts, hosting, and web, iOS, and Android publishing. Product Hunt ranked Floot MCP second on September 24, 2026.
Sources: Product Hunt, Floot on Product Hunt
Solid launched a new Product Hunt offering in which agents can build apps, operate software through APIs or interfaces, provision tools, and spend within configured budgets. Product Hunt reported Solid ranked first for the day on September 23, 2026; Solid describes support for long-running app-building and workflow tasks.
Sources: Product Hunt, Solid
GitHub made new Copilot code-review settings generally available, including automatic reviews for pull-request events, configurable review effort, and enterprise-wide default effort settings. The controls are available across Copilot plans, including Business and Enterprise.
Sources: GitHub Changelog
GitHub added project-level sandbox controls for Copilot app sessions, covering filesystem access, network access, and Git or GitHub CLI credentials. The feature is off by default and available in public preview.
Sources: GitHub Changelog
GitHub will remove RSA SHA-1 signatures and an older Diffie-Hellman exchange method, require new RSA SSH keys to be at least 3072 bits from October 14, 2026, and add ML-KEM-based key exchange outside the U.S. region.
Sources: GitHub Changelog
Copilot CLI now supports whole-codebase indexing for C++ projects, creating a reusable symbol index across source files and headers to speed definitions, references, implementations, and symbol searches.
Sources: GitHub Changelog
GitHub made Anthropic Claude Opus 5.5 and OpenAI GPT-6 Sol and GPT-6 Luna available across Copilot clients and agent experiences, with access varying by plan and model rollout.
Sources: GitHub Changelog, GitHub Changelog
Copilot for JetBrains 1.18.0 adds public-preview assisted approvals for low-risk tool calls, organization-shared skills and instructions, Codex plan mode, and persistent controls for MCP tools.
Sources: GitHub Changelog
GitHub Copilot app administrators can now export agent-session telemetry through OpenTelemetry, including model requests, tool usage, and execution traces. Prompt and response content is excluded by default.
Sources: GitHub Changelog
Hyrax AI launched an agent that audits entire repositories across security, correctness, performance, maintainability, architecture, and operations, then proposes verified fixes through pull requests. Product Hunt ranked it fifth for September 21, 2026.
Sources: Product Hunt, Hyrax AI
GitHub Enterprise Cloud now lets enterprise owners export inventories of credentials including SSH keys, personal access tokens, OAuth tokens, and GitHub App tokens. Exports include ownership, permissions, creation and expiration dates, usage data, and organization or repository targets to support incident response.
Sources: GitHub Changelog
Vybe announced a $10 million seed round led by First Round Capital, with participation from Y Combinator and operators. The company is building a vibe-coding platform for internal applications with production-data integrations, SSO, access controls, Git workflows, and a security layer that AI cannot modify.
Sources: ContentEdge, Vybe