# X Stack

generated_at: 2026-08-29T22:31:14+00:00
published_at: 2026-08-29T22:32:04+00:00

Seed catalog from official SpaceXAI, Cursor, and X sources. Channel pointers (stable/alpha/enterprise) read from x.ai/cli on 2026-08-22. Listing/sitemap pages are not scrape sources.

Products: 10
Playbooks: 10

## Grok 4.6 (`grok-4-6`)

- org: spacexai
- status: ga
- current_version: 4.6
- last_updated: 2026-08-25
- tagline: Frontier model for long-running agents and ambitious visual work.
- product: https://x.ai/news/grok-4-6
- docs: https://docs.x.ai/developers/grok-4-6
- changelog: https://docs.x.ai/developers/release-notes
- tags: model, agents, coding, api, long-context

Grok 4.6 builds on Grok 4.5 with a focus on staying with complex tasks across many steps — research, analysis, codebase work, and turning an idea into a polished application. 500k context, text and image inputs, text-only output, and no text output limit. Available in Cursor, Grok Build, the xAI API, and partners including OpenRouter, Vercel, Cloudflare, Amazon Bedrock, and GitHub Copilot.

### Channels

No documented install channels.

### Features

- **Long-running agents** (`grok-4-6-long-running-agents`) `agents`, `multi-agent`, `computer-use`
  Stays with multi-step work: research, analysis, codebase edits, and turning a product idea into a working first version.
  - **Self-testing and verification** (`grok-4-6-self-testing`) `agents`, `coding`
    On longer trajectories, checks its own work before moving on.
- **Interactive and visual work** (`grok-4-6-visual-interactive`) `ui`, `coding`
  Stronger first passes on visual and interactive projects — structure and visual language in one pass.
- **500k context window** (`grok-4-6-context`) `api`, `long-context`
  Text and image inputs with text-only output and no text output limit.
- **Reasoning effort** (`grok-4-6-reasoning-effort`) `api`, `reasoning`
  Configurable reasoning: low, medium, high (default), and xhigh.
- **API pricing** (`grok-4-6-pricing`) `api`, `pricing`
  $2 / $0.50 / $6 per 1M tokens (input / cached / output) below 200k prompt tokens; $4 / $1 / $12 above. Fast variant is twice the price.
- **Where it runs** (`grok-4-6-availability`) `api`, `availability`
  Cursor, Grok Build, xAI API, OpenRouter, Vercel, Cloudflare, Amazon Bedrock, GitHub Copilot, Gemini Enterprise Agent Platform.

### Changelog

- **2026-08-21** · availability · Grok 4.6 on Gemini Enterprise Agent Platform (`grok-4-6-vertex-ai-2026-08-21`, high)
  Grok 4.6 is available on Gemini Enterprise Agent Platform via Model Garden, with 500k context and configurable reasoning (low, medium, high, xhigh).
- **2026-08-19** · availability · Grok 4.6 on Amazon Bedrock (`grok-4-6-bedrock-2026-08-19`, high)
  Flagship model available on Amazon Bedrock with 500k context and configurable reasoning levels.
- **2026-08-14** · availability · Grok 4.6 in GitHub Copilot (`grok-4-6-copilot-2026-08-14`, high)
  Grok 4.6 listed as available inside GitHub Copilot.
- **2026-08-12** · release · Introducing Grok 4.6 (`grok-4-6-release-2026-08-12`, high)
  Released for long-running agents and interactive/visual work. Live in Cursor, Grok Build, and the API the same day, with 2x included usage in Cursor and Grok Build for the first week.

## Grok Bot (`grok-bot`)

- org: spacexai
- status: beta
- current_version: beta
- last_updated: 2026-08-26
- tagline: Always-on AI teammates with their own cloud computer.
- product: https://x.ai/news/introducing-grok-bot
- docs: https://docs.x.ai/grok-bot/overview
- tags: agents, computer-use, multi-agent, always-on

Grok Bot is an early-beta team of always-on agents. Each bot has a cloud computer, can sign into apps and websites (including tools with no clean API or MCP), message like a colleague, and keep working after you close the laptop. Bots can message each other, coordinate in group chats, save workflows as routines, and only pull you in for approvals. Included with SuperGrok Plus, SuperGrok Heavy, Cursor Pro+, Cursor Ultra, and Cursor Teams (Standard and Premium) on desktop and iOS.

### Channels

No documented install channels.

### Features

- **Own cloud computer** (`grok-bot-computer`) `computer-use`, `agents`
  Each Bot runs on a persistent cloud VM with a browser, filesystem, and terminal. All Bots on an account share that user-scoped computer (files, browser sessions, logins), with a separate screen per Bot.
- **Always-on runtime** (`grok-bot-always-on`) `always-on`, `agents`
  Keep working 24/7 in the cloud after the client is closed.
- **Colleague messaging** (`grok-bot-messaging`) `multi-agent`, `messaging`
  Message a bot like a teammate. Bots remember conversations, share context in threads, and can coordinate in group chat.
- **Skills** (`grok-bot-skills`) `skills`, `automation`
  A skill is a reusable set of instructions for how to do a task. Skills are available across Bots. Teach a task can record a browser workflow (up to ten minutes) into a draft skill when that control is visible.
- **Routines** (`grok-bot-routines`) `automation`, `agents`
  A routine tells one Bot when to run a workflow — on a schedule or, where supported, after an event. Background routines can run while the laptop is closed. A Bot can own up to 50 routines.
- **Supported platforms** (`grok-bot-platforms`) `availability`
  Desktop app on macOS (Apple silicon and Intel) and Windows (x64 and Arm64); companion app on iPhone (iOS 18 or later). Linux desktop, Android, and iPad are not supported at initial launch.
- **Approvals** (`grok-bot-approvals`) `agents`
  Bots only pull you in for judgment calls. They follow up on dropped threads and nudge stalled handoffs.
- **Connectors** (`grok-bot-connectors`) `connectors`, `computer-use`
  Sign into the tools you already use and work across apps and inboxes, including platforms with no clean API or MCP.

### Changelog

- **2026-08-21** · availability · Grok Bot is now included with more plans (`grok-bot-more-plans-2026-08-21`, high)
  Grok Bot is included with SuperGrok Plus, SuperGrok Heavy, Cursor Pro+, Cursor Ultra, and Cursor Teams (Standard and Premium).
- **2026-08-11** · release · Introducing Grok Bot (`grok-bot-launch-2026-08-11`, high)
  Early beta of always-on AI teammates with a shared cloud computer. Available to SuperGrok Heavy, Cursor Ultra, and Cursor Teams Premium on desktop and iOS; enterprise waitlist.

## Grok Build (`grok-build`)

- org: spacexai
- status: ga
- current_version: 1.0.13
- last_updated: 2026-08-29
- tagline: Coding agent for plan, build, test, and deploy — now on Grok 4.6.
- product: https://x.ai/build
- docs: https://docs.x.ai/build/overview
- changelog: https://x.ai/build/changelog
- tags: coding, agents, cli, mcp

Grok Build is SpaceXAI’s coding agent for professional software engineering. Plan mode blocks edits until approved. Subagents run in parallel with their own context. Skills, hooks, MCP servers, and AGENTS.md adapt it to a repo. Plugins and marketplaces extend it; Claude Code instruction files are read with no extra setup. Use the TUI, headless grok -p, or ACP. Stable CLI is 1.0.13; alpha and enterprise channels also point to 1.0.13 (GROK_CHANNEL). Powered by Grok 4.6; also shipping on web and mobile.

### Channels

- `stable` · 1.0.13 · updated 2026-08-29
  - install: `curl -fsSL https://x.ai/cli/install.sh | bash`
  - notes: Default. Channel pointer https://x.ai/cli/stable currently serves 1.0.13. Changelog on x.ai/build/changelog is not a scrape source; feature notes for 1.0.13 have not been cited from a 200 page.
- `alpha` · 1.0.13 · updated 2026-08-28
  - install: `curl -fsSL https://x.ai/cli/install.sh | GROK_CHANNEL=alpha bash`
  - notes: Pre-release binaries. Channel pointer https://x.ai/cli/alpha currently serves 1.0.13. Changelog on x.ai/build/changelog is not a scrape source; feature notes for 1.0.13 have not been cited from a 200 page.
- `enterprise` · 1.0.13 · updated 2026-08-28
  - install: `curl -fsSL https://x.ai/cli/install.sh | GROK_CHANNEL=enterprise bash`
  - notes: Enterprise channel pointer https://x.ai/cli/enterprise currently serves 1.0.13. Changelog on x.ai/build/changelog is not a scrape source; feature notes for 1.0.13 have not been cited from a 200 page.

### Features

- **Plan mode** (`grok-build-plan-mode`) `coding`, `agents`
  The agent explores and drafts a plan before edits. Only the session plan file may be edited until you approve; bash is not gated. Auto and always-approve do not skip plan review. Enter with /plan or Shift+Tab; the agent can enter plan mode on its own. Reopen with /view-plan.
- **Auto and always-approve** (`grok-build-permission-modes`) `cli`, `agents`
  Ask (default) prompts for anything not already allowed. Auto uses a classifier to auto-approve safe tools. Always-approve skips permission prompts; deny rules and PreToolUse hooks still apply. Shift+Tab cycles Normal → Plan → Auto → Always-approve. Allow/deny rules in [permission] or --allow/--deny; deny wins.
- **Headless and scripting** (`grok-build-headless`) `cli`, `automation`
  Run without the TUI via grok -p. Optional --output-format streaming-json for scripts, automations, or other apps.
- **Agent Client Protocol** (`grok-build-acp`) `cli`, `agents`
  Use Grok Build through the Agent Client Protocol (ACP) in other apps.
- **Subagents** (`grok-build-subagents`) `multi-agent`, `coding`
  Independent child sessions with their own context; they return a summary to the parent. Built-in types: general-purpose, explore (read/list/search only), and plan (no shell, no edits). Add types under .grok/agents/ or ~/.grok/agents/. Personas are behavioral overlays via /personas.
- **Skills** (`grok-build-skills`) `skills`, `coding`
  Reusable folders with markdown, scripts, and resources (SKILL.md). Discovered from ./.grok/skills/, ~/.grok/skills/, plugin skills/, and [skills] paths. User-invocable skills appear as /<skill-name>. Capture a session as a skill with /skillify.
- **Plugins** (`grok-build-plugins`) `skills`, `mcp`, `cli`
  Plugins add skills, agents, hooks, MCP servers, and LSP servers from ./.grok/plugins/, ~/.grok/plugins/, marketplace installs, [plugins] paths, or --plugin-dir. The TUI extensions modal opens from /plugins, /hooks, /skills, or /mcps.
- **Marketplaces** (`grok-build-marketplaces`) `skills`, `cli`
  TUI Marketplace tab installs plugins from [[marketplace.sources]] in ~/.grok/config.toml and ~/.grok/plugins/known_marketplaces.json.
- **Claude Code compatibility** (`grok-build-claude-compat`) `coding`, `skills`
  Reads Claude Code marketplaces, plugins, skills, MCPs, agents, hooks, and instruction files (CLAUDE.md, Claude.md, CLAUDE.local.md, .claude/rules/) alongside .grok/ with no extra setup.
- **Hooks** (`grok-build-hooks`) `hooks`, `coding`
  JSON hooks (command or HTTP) on lifecycle events. PreToolUse is the only blocking event (explicit deny, else fail-open). Personal ~/.grok/hooks/*.json; project .grok/hooks/*.json needs /hooks-trust or --trust. Also reads Claude Code and Cursor hook files.
- **MCP servers** (`grok-build-mcp`) `mcp`, `connectors`
  MCP servers expose external tools as <server>__<tool>. Add with grok mcp add (stdio or HTTP/OAuth). Also declared in ~/.grok/config.toml or project --scope. Loads ~/.claude.json, .cursor/mcp.json, and .mcp.json unless compat is disabled. /mcps in the TUI.
- **AGENTS.md** (`grok-build-agents-md`) `coding`
  Loads AGENTS.md, Agents.md, AGENT.md, CLAUDE.md, and *.md in .grok/rules/ (plus .claude/rules/ and .cursor/rules/) from repo root down to cwd; deeper files win. --rules appends session-only text. grok inspect lists loaded files.
- **Web and mobile** (`grok-build-web-mobile`) `availability`
  Grok Build is available beyond the terminal — web and mobile surfaces announced Aug 19, 2026.
- **Release channels** (`grok-build-channels`) `cli`, `alpha`
  Installer respects GROK_CHANNEL=stable|alpha|enterprise (default stable). The welcome screen shows the active channel. Failed channel updates suggest the matching reinstall command. Persist the choice in ~/.grok/config.toml under [cli] channel.
- **Sessions** (`grok-build-sessions`) `cli`, `agents`
  Every conversation is saved under ~/.grok/sessions/. Resume with /resume, grok --resume, or grok -c. /fork branches a peer session; /rewind restores files and truncates the conversation. grok sessions list/search/delete and grok export manage them.
- **Worktrees** (`grok-build-worktrees`) `cli`, `git`
  Isolated git checkouts under ~/.grok/worktrees/<repo>/<name> so parallel agents do not overwrite each other. Start with grok -w, /fork --worktree, or Ctrl+W on the dashboard. Persist until grok worktree rm or gc.
- **Sandbox** (`grok-build-sandbox`) `cli`, `infra`
  Limits what an approved tool call can read, write, and reach on the network (Landlock on Linux, Seatbelt on macOS). Off by default. Profiles: off, workspace, devbox, read-only, strict. Enable with grok --sandbox, [sandbox] profile, or GROK_SANDBOX.
- **Agent Dashboard** (`grok-build-dashboard`) `cli`, `agents`
  Fullscreen overview of every session (needs input, working, idle, completed, failed). Open with Ctrl+\, /dashboard, or grok dashboard. Reply or answer permission prompts inline; Ctrl+W starts new agents in a worktree.
- **Background tasks** (`grok-build-background-tasks`) `cli`, `automation`
  Run commands, subagents, and monitors while the conversation continues (/tasks, Ctrl+G). /loop repeats a prompt on an interval (max 50, expire after 7 days). Monitors notify on each script line. Ctrl+; or /queue lists prompts queued behind the running turn.
- **Theming** (`grok-build-theming`) `cli`
  TUI themes via /theme or [ui] theme in ~/.grok/config.toml. Built-ins include GrokNight (default), GrokDay, TokyoNight, RosePineMoon, and OscuraMidnight. theme = auto follows OS light/dark.

### Changelog

- **2026-08-29** · release · stable · Stable channel 1.0.13 (`grok-build-stable-1-0-13-2026-08-29`, high)
  Official stable pointer https://x.ai/cli/stable currently serves 1.0.13. Alpha and enterprise pointers also serve 1.0.13. Feature notes for 1.0.13 have not yet landed on the public changelog page.
- **2026-08-28** · release · enterprise · Enterprise channel 1.0.13 (`grok-build-enterprise-1-0-13-2026-08-28`, high)
  Official enterprise pointer https://x.ai/cli/enterprise currently serves 1.0.13. Stable remains 1.0.5; alpha pointer also serves 1.0.13. Feature notes for 1.0.13 have not yet landed on the public changelog page.
- **2026-08-28** · release · alpha · Alpha channel 1.0.13 (`grok-build-alpha-1-0-13-2026-08-28`, high)
  Official alpha pointer https://x.ai/cli/alpha currently serves 1.0.13. Stable remains 1.0.5; enterprise pointer also serves 1.0.13. Feature notes for 1.0.13 have not yet landed on the public changelog page.
- **2026-08-27** · release · enterprise · Enterprise channel 1.0.12 (`grok-build-enterprise-1-0-12-2026-08-27`, high)
  Official enterprise pointer https://x.ai/cli/enterprise currently serves 1.0.12. Stable remains 1.0.5; alpha pointer also serves 1.0.12. Feature notes for 1.0.12 have not yet landed on the public changelog page.
- **2026-08-27** · release · alpha · Alpha channel 1.0.12 (`grok-build-alpha-1-0-12-2026-08-27`, high)
  Official alpha pointer https://x.ai/cli/alpha currently serves 1.0.12. Stable remains 1.0.5; enterprise pointer remains 1.0.11. Feature notes for 1.0.12 have not yet landed on the public changelog page.
- **2026-08-26** · release · alpha · Alpha channel 1.0.11 (`grok-build-alpha-1-0-11-2026-08-26`, high)
  Official alpha pointer https://x.ai/cli/alpha currently serves 1.0.11. Stable remains 1.0.5; enterprise pointer now also serves 1.0.11. Feature notes for 1.0.11 have not yet landed on the public changelog page.
- **2026-08-26** · release · enterprise · Enterprise channel 1.0.11 (`grok-build-enterprise-1-0-11-2026-08-26`, high)
  Official enterprise pointer https://x.ai/cli/enterprise currently serves 1.0.11. Stable remains 1.0.5; alpha pointer also serves 1.0.11. Feature notes for 1.0.11 have not yet landed on the public changelog page.
- **2026-08-25** · release · alpha · Alpha channel 1.0.10 (`grok-build-alpha-1-0-10-2026-08-25`, high)
  Official alpha pointer https://x.ai/cli/alpha currently serves 1.0.10. Stable remains 1.0.5; enterprise pointer remains 1.0.8. Feature notes for 1.0.10 have not yet landed on the public changelog page.
- **2026-08-24** · release · alpha · Alpha 1.0.9 (superseded) (`grok-build-alpha-1-0-9-2026-08-24`, medium)
  Alpha briefly served 1.0.9. Binaries are published at https://x.ai/cli/grok-1.0.9-linux-x86_64. Alpha now points at 1.0.10. Feature notes for 1.0.9 have not landed on the public changelog page.
- **2026-08-20** · release · alpha · Alpha channel 1.0.8 (`grok-build-alpha-1-0-8-2026-08-20`, high)
  Official alpha pointer https://x.ai/cli/alpha served 1.0.8 (enterprise pointer still matches). Stable remains 1.0.5. Alpha now points at 1.0.10. Feature notes for 1.0.8 have not landed on the public changelog page.
- **2026-08-19** · availability · Grok Build on web and mobile (`grok-build-web-mobile-2026-08-19`, high)
  Grok Build is available on every plan, on the web and on mobile.
- **2026-08-19** · release · alpha · Alpha 1.0.6 (superseded) (`grok-build-alpha-1-0-6-2026-08-19`, medium)
  Alpha briefly served 1.0.6: Shift+arrow prompt selection, optional status line, grok clone with a projected working tree, plus reliability fixes. Binaries were published at https://x.ai/cli/grok-1.0.6-*. Alpha now points at 1.0.10.
- **2026-08-15** · release · stable · Grok Build 1.0.5 (`grok-build-1-0-5-2026-08-15`, high)
  GROK_CONFIG and GROK_CONFIG_PATH let launchers override config without editing config.toml. Worktrees under ~/.grok/worktrees are reclaimed when safe.
- **2026-08-13** · release · stable · Grok Build 1.0.4 (`grok-build-1-0-4-2026-08-13`, high)
  Recurring /loop tasks show a 7-day expiry notice. Web search allow/exclude domains via [toolset.web_search]. GROK_SESSION_ID for tools and MCP. Follow-up Behavior setting, session search toggle, and many TUI/permission fixes.
- **2026-08-12** · feature · Powered by Grok 4.6 (`grok-build-4-6-2026-08-12`, high)
  Grok Build switched to Grok 4.6 on launch day, with 2x included usage for the first week.
- **2026-08-12** · release · stable · Grok Build 1.0.3 (`grok-build-1-0-3-2026-08-12`, high)
  Clickable /session-info rows, faster subagent spawning with many ~/.grok sessions, and TUI rendering matched to 120 Hz+ displays.
- **2026-08-11** · release · stable · Grok Build 1.0.2 (`grok-build-1-0-2-2026-08-11`, high)
  Clearer tool-call streaming labels, recovery from server-rejected images, actionable startup timeout messages, safer worktree copies of large repos, and Cmd+click autolink fixes in Apple Terminal.
- **2026-08-10** · release · stable · Grok Build 1.0.1 (`grok-build-1-0-1-2026-08-10`, high)
  /rewind truncates conversation only. grok du for disk usage. Failed channel updates suggest the matching GROK_CHANNEL reinstall. Native arm64 install on Apple Silicon even from a Rosetta shell. Video generation voices and aspect ratios.
- **2026-08-07** · release · stable · Grok Build 1.0.0 (`grok-build-1-0-0-2026-08-07`, high)
  1.0: dashboard turn summaries, extensions modal, auto theme over SSH/tmux, MCP image tools, cleaner API error banners, and a large set of TUI and permission-prompt fixes.
- **2026-07-28** · feature · Introducing Build Mode (`grok-build-build-mode-2026-07-28`, high)
  Build Mode Early Beta for SuperGrok Heavy on grok.com, iOS, and Android: describe an idea and Grok builds a working preview in chat, then publish it to a link.
- **2026-07-23** · feature · Workflows in Grok Build (`grok-build-workflows-2026-07-23`, high)
  Grok Build can write and run workflows: orchestration scripts that fan a task out across many agents.
- **2026-07-15** · availability · Grok Build is Now Open Source (`grok-build-open-source-2026-07-15`, high)
  Grok Build coding agent and TUI source is on GitHub; the harness can run local-first against your own inference.

## Grok Imagine (`grok-imagine`)

- org: spacexai
- status: ga
- current_version: image-2.0 / video-1.5
- last_updated: 2026-08-07
- tagline: Image and video generation and editing for real creative work.
- product: https://grok.com/imagine
- docs: https://docs.x.ai/developers/model-capabilities/imagine
- changelog: https://x.ai/news/grok-imagine-image-2
- tags: media, api, image, video

Imagine covers stills and motion. Image 2.0 (Aug 7, 2026) is Quality Mode on grok.com/imagine and the iOS/Android apps, and grok-imagine-image-2.0 in the API. It follows layout and typography closely, edits with up to 5 reference images, and includes magic wand, segmentation, background removal, and smart resize. Video 1.5 supports text-to-video, image-to-video, and reference-to-video, including native 1080p for T2V/I2V.

### Channels

No documented install channels.

### Features

- **Imagine Image 2.0** (`grok-imagine-image-2`) `image`, `api`
  Quality Mode for precise generation and editing. API id grok-imagine-image-2.0. 1K and 2K stills.
- **First-class editing** (`grok-imagine-editing`) `image`
  Magic wand, segmentation, background removal, multi-ref editing (up to 5 inputs), smart resize across common aspect ratios.
- **Imagine Video 1.5** (`grok-imagine-video-1-5`) `video`, `api`
  Text-to-video, image-to-video, and reference-to-video. Native 1080p for T2V and I2V (Jul 31, 2026).
- **Imagine API** (`grok-imagine-api`) `api`, `media`
  Generate and edit images and video through the xAI API. Image input billed separately from output.

### Changelog

- **2026-08-07** · release · Imagine Image 2.0 (`grok-imagine-image-2-2026-08-07`, high)
  Generally available as Quality Mode on grok.com/imagine, iOS, and Android. API model grok-imagine-image-2.0.
- **2026-07-31** · feature · Imagine Video 1.5 with References (`grok-imagine-video-1-5-2026-07-31`, high)
  Image and voice references, text-to-video, and native 1080p. API model grok-imagine-video-1.5.

## Grok Voice (`grok-voice`)

- org: spacexai
- status: ga
- current_version: think-fast-2.0
- last_updated: 2026-08-26
- tagline: Realtime speech, TTS, and STT through one Voice API.
- product: https://x.ai/api
- docs: https://docs.x.ai/developers/model-capabilities/audio/voice
- tags: voice, api, realtime

The Voice API covers speech-to-speech, text-to-speech, and speech-to-text. Voice Think Fast 2.0 was introduced Jul 29, 2026. Speech-to-speech starts at $0.05 / min; TTS is $15.00 / 1M characters. STT gained an adjustable VAD threshold on Jul 23, 2026.

### Channels

No documented install channels.

### Features

- **Speech to speech** (`grok-voice-s2s`) `voice`, `api`, `realtime`
  Realtime voice agents. Starting at $0.05 / min.
- **Text to speech** (`grok-voice-tts`) `voice`, `api`
  TTS at $15.00 / 1M characters.
- **Speech to text** (`grok-voice-stt`) `voice`, `api`
  STT with adjustable VAD threshold (vad_threshold) for streaming and batch.
- **Voice Think Fast 2.0** (`grok-voice-think-fast-2`) `voice`
  Introduced Jul 29, 2026. grok-voice-latest routes to grok-voice-think-fast-2.0 starting August 5, 2026.
- **Custom voices** (`grok-voice-custom-voices`) `voice`, `api`
  Clone a voice from a reference clip (max 120s) via /v1/custom-voices and use the returned voice_id for TTS and speech-to-speech.

### Changelog

- **2026-07-29** · release · Introducing Grok Voice Think Fast 2.0 (`grok-voice-think-fast-2-2026-07-29`, high)
  Grok Voice Think Fast 2.0 announced as the next-generation speech-to-speech voice model.
- **2026-07-23** · feature · Adjustable VAD threshold (`grok-voice-vad-2026-07-23`, high)
  Speech-to-text accepts vad_threshold as a streaming query param and batch multipart field. 0 disables the gate.
- **2026-07-06** · feature · 21 New Flagship Grok Voices (`grok-voice-flagship-voices-2026-07-06`, high)
  21 new multilingual flagship voices for Grok Voice, plus improved naturalness for the original five, in the Voice Agent API, TTS API, and Voice Agent Builder.
- **2026-07-01** · feature · Introducing the Voice Agent Builder (`grok-voice-agent-builder-2026-07-01`, high)
  Voice Agent Builder beta: no-code production voice agents on Grok Voice, with telephony, tools, MCP, and observability.

## Grok API / Console (`grok-api`)

- org: spacexai
- status: ga
- current_version: 2026-08
- last_updated: 2026-08-19
- tagline: One API for text, images, video, voice, and search.
- product: https://console.x.ai
- docs: https://docs.x.ai/developers/release-notes
- changelog: https://docs.x.ai/developers/release-notes
- tags: api, console, model

The xAI API and console at console.x.ai expose Grok models, Imagine, and Voice. Grok 4.6 is the current flagship for coding, agentic tasks, and knowledge work: 500k context, reasoning effort including xhigh, cached-input pricing, and no text output limit. Partners include OpenRouter, Vercel, Cloudflare, Amazon Bedrock, and GitHub Copilot.

### Channels

No documented install channels.

### Features

- **Flagship models** (`grok-api-models`) `api`, `model`
  Grok 4.6 for agents, coding, and knowledge work. Text and image in, text out.
- **Media and voice** (`grok-api-media-voice`) `api`, `media`, `voice`
  Imagine image/video generation and editing plus Voice realtime, TTS, and STT on the same API.
- **Console** (`grok-api-console`) `console`, `api`
  Keys, playground, usage, and changelog at console.x.ai.
- **Partner surfaces** (`grok-api-partners`) `availability`, `api`
  OpenRouter, Vercel, Cloudflare, Amazon Bedrock, GitHub Copilot.

### Changelog

- **2026-08-12** · release · Grok 4.6 on the xAI API (`grok-api-4-6-2026-08-12`, high)
  500k context; $2 / $0.50 / $6 per 1M tokens below 200k prompt tokens, $4 / $1 / $12 above; reasoning low / medium / high / xhigh.

## Cursor Cloud Agents (`cursor-cloud-agents`)

- org: cursor
- status: ga
- current_version: 2026-08-27
- last_updated: 2026-08-27
- tagline: Always-on agents that pick up work from events and hold a goal.
- product: https://cursor.com
- docs: https://cursor.com/changelog
- changelog: https://cursor.com/changelog/start-from-scratch
- tags: agents, always-on, multi-agent, automation

Cloud agents run in Cursor’s harness so they can operate as a system: subscribe to PRs, Slack, or a schedule; hold a /goal until it is met; spawn subagents on isolated VMs; and take follow-up steering without interrupting the current turn. Builds pre-warm the environment so agents start ~3× faster. Hooks observe prompts, responses, thinking, subagents, and turn completion.

### Channels

No documented install channels.

### Features

- **Subscriptions** (`cursor-ca-subscriptions`) `automation`, `always-on`
  Monitor PRs, watch a Slack thread, or run scheduled tasks. Agents wake on events and can auto-subscribe to PRs they create.
- **/goal and /loop** (`cursor-ca-goal`) `agents`, `automation`
  Assign a long-lived objective (e.g. fix flaky tests and make CI green). Pair with /loop for recurring check-ins.
- **Subagents on their own machines** (`cursor-ca-subagents`) `multi-agent`, `computer-use`
  Subagents run on isolated VMs with clean project copies. Swarm them for independent tests.
- **Custom modes** (`cursor-ca-custom-modes`) `skills`, `agents`
  Pin skills as always-on modes so agents stay focused. Select from the / menu.
- **Steering** (`cursor-ca-steering`) `agents`
  Send a follow-up without interrupting the current turn. Press return twice or Send now.
- **Builds** (`cursor-ca-builds`) `agents`, `infra`
  Ready-to-use copies of the dev environment. Agents boot ~3× faster to first token; included with Cloud Agents.
- **Conversation hooks** (`cursor-ca-hooks`) `hooks`, `agents`
  Observe and control prompts, responses, thinking, subagents, compaction, and turn completion.
- **Start from scratch** (`cursor-ca-start-from-scratch`) `agents`, `git`
  Cloud Agents can start without a connected GitHub or other SCM. Cursor creates an Origin repo in the background; save with Create repo when ready.
- **Live preview** (`cursor-ca-live-preview`) `agents`, `computer-use`
  Port-forwards the cloud agent environment to the browser for preview and design mode.
- **Vercel publish** (`cursor-ca-vercel-publish`) `agents`, `connectors`
  Connect a Vercel account and publish the agent build to a live URL.

### Changelog

- **2026-08-27** · feature · Start from scratch, without a repo (`cursor-ca-start-from-scratch-2026-08-27`, high)
  Cloud Agents can start without a connected GitHub or other SCM. Select Start from scratch in the repo picker; Cursor creates an Origin repo in the background. Create repo saves private or internal Origin repos. Live preview port-forwards the agent environment to the browser. Connect Vercel to publish a live URL.
- **2026-08-19** · feature · Cloud Agents and harness improvements (`cursor-ca-harness-2026-08-19`, high)
  Subscriptions, custom modes, subagents on isolated VMs, /goal, steering without interruption, and expanded conversation hooks.
- **2026-08-13** · feature · Cloud Agents start 3× faster with Builds (`cursor-ca-builds-2026-08-13`, high)
  Background-prepared environments (clone, install, install script). Fallback to last successful build. Builds tab for logs and SHAs.

## Cursor Origin (`cursor-origin`)

- org: cursor
- status: beta
- current_version: early-beta
- last_updated: 2026-08-17
- tagline: Cursor can host your code.
- product: https://cursor.com/changelog/origin-code-hosting
- changelog: https://cursor.com/changelog/origin-code-hosting
- tags: git, agents, hosting

Origin is early-beta code hosting on paid plans. The Codebase tab holds Origin repos, GitHub sync, pull requests, and Apps (Vercel previews, Depot/Buildkite CI). Agents can query code, change files, update PRs, and push branches from every repo. Designed for agent scale; agent-native features continue to ship.

### Channels

No documented install channels.

### Features

- **Origin repos** (`cursor-origin-repos`) `git`, `hosting`
  Create and host repos from the Codebase tab. Clone and push with the CLI. URL form cursor.com/codebase/{name}.
- **GitHub sync** (`cursor-origin-github`) `git`
  Bidirectional PR sync and real-time updates. Disconnect anytime.
- **Pull requests** (`cursor-origin-prs`) `git`, `agents`
  Timeline, commits, checks, diffs, comments, and merges in Origin, two-way synced with GitHub.
- **Agents in every repo** (`cursor-origin-agents`) `agents`, `git`
  Query code, make changes, update PRs, and push branches from hosted repos.
- **App extensions** (`cursor-origin-apps`) `connectors`
  Vercel PR previews, Depot/Buildkite CI. Connect from the Apps tab.

### Changelog

- **2026-08-17** · release · Origin code hosting (`cursor-origin-launch-2026-08-17`, high)
  Early beta on paid plans: repos, GitHub sync, pull requests, agents in every repo, and app extensions. Agent-native features still shipping.

## Cursor IDE / CLI (`cursor-desktop`)

- org: cursor
- status: ga
- current_version: 3.17
- last_updated: 2026-08-21
- tagline: Desktop and CLI surface for Grok 4.6, cloud agents, and Origin.
- product: https://cursor.com/download
- changelog: https://cursor.com/changelog
- tags: ide, cli, coding

Cursor’s desktop IDE and CLI are the local control plane for Cloud Agents, Origin, and Grok 4.6. Current stable desktop on cursor.com/download is 3.17. Pre-release builds roll through Early Access and Nightly (Settings → Beta → Update Access) before the default track. Feature write-ups on the changelog are per minor, not per patch build.

### Channels

- `stable` · 3.17 · updated 2026-08-21
  - install: `https://cursor.com/download`
  - notes: Download page lists 3.17 as Latest for macOS, Windows, and Linux. Patch builds (3.17.x) are not written up individually. cursor.com/changelog has no 3.17 heading.
- `early-access` · pre-release · updated 2026-08-20
  - notes: Settings → Beta → Update Access. Desktop minors often land on Early Access before the default track.
- `nightly` · nightly · updated 2026-08-20
  - notes: Settings → Beta → Update Access → Nightly. No public version pointer; Cursor does not publish a separate nightly changelog.

### Features

- **Grok 4.6 in Cursor** (`cursor-desktop-grok-4-6`) `model`, `coding`
  Flagship model available in Cursor on Aug 12, 2026 with a one-week 2× usage boost.
- **CLI** (`cursor-desktop-cli`) `cli`, `git`
  Clone and push Origin repos, drive agents, and script workflows from the terminal.
- **Google Workspace plugins** (`cursor-desktop-plugins`) `connectors`
  Gmail, Drive, and Calendar access from Cursor (Aug 3, 2026 changelog).
- **iPad app** (`cursor-desktop-ipad`) `availability`
  Inbox, PR reviews, and multi-PR sessions on iPad.
- **Early Access and Nightly** (`cursor-desktop-prerelease`) `alpha`, `ide`
  Pre-release desktop builds via Settings → Beta → Update Access. Feature changelog entries ship with the minor; patch builds are not listed separately.

### Changelog

- **2026-08-21** · release · stable · Cursor 3.17 Latest (`cursor-desktop-3-17-2026-08-21`, high)
  cursor.com/download lists 3.17 as the current desktop for macOS, Windows, and Linux. Feature notes have not landed as a per-patch write-up on cursor.com/changelog (latest product notes remain Cloud Agents, Origin, Builds).
- **2026-08-18** · release · stable · Cursor 3.16 Latest (`cursor-desktop-3-16-2026-08-18`, high)
  cursor.com/download lists 3.16 as the current desktop for macOS, Windows, and Linux. Feature notes remain on the product changelog (Cloud Agents, Origin, Builds) rather than per patch.
- **2026-08-12** · availability · Grok 4.6 in Cursor (`cursor-desktop-4-6-2026-08-12`, high)
  Grok 4.6 available in Cursor with 2× included usage for the first week.
- **2026-08-03** · feature · Google Workspace plugins (`cursor-desktop-workspace-2026-08-03`, high)
  Gmail, Drive, and Calendar plugins listed on the Cursor changelog.

## X API (`x-api`)

- org: x
- status: ga
- current_version: v2
- last_updated: 2026-08-13
- tagline: Pay-per-usage X API v2, Activity API, and media.
- product: https://docs.x.com/x-api/getting-started/pricing
- docs: https://docs.x.com/x-api/activity/introduction
- changelog: https://docs.x.com/changelog
- tags: api, activity, media, pricing

X API v2 uses pay-per-usage credits rather than subscriptions. The X Activity API (XAA) delivers realtime user activity; inbound webhook events are billed per delivery with 24-hour UTC dedup. Recent additions include mute/block events (Jul 21, 2026) and post create/delete (Jun 4, 2026). Media upload remains on the v1.1 media endpoints. Owned reads are priced separately from public lookup.

### Channels

No documented install channels.

### Features

- **X Activity API** (`x-api-activity`) `api`, `activity`
  Realtime user-activity webhooks. Successor path from legacy Account Activity API. Subscribe by user id; optional direction filter.
- **Pay-per-usage credits** (`x-api-ppu`) `api`, `pricing`
  No subscription SKUs. Buy credits, set spending limits, earn up to 20% back in xAI API credits. Owned reads vs public lookup priced separately.
- **Webhook event billing** (`x-api-webhooks-billing`) `api`, `pricing`, `activity`
  Activity API events billed per delivery, same 24h UTC dedup window. Inbound events cost; many outbound actions do not.
- **Media upload** (`x-api-media`) `media`, `api`
  Media upload remains on the v1.1 media endpoints after other v1.1 SKUs were retired.
- **Mute and block events** (`x-api-mute-block`) `activity`, `api`
  mute.mute / mute.unmute / block.block / block.unblock. Private events; OAuth 2.0 user-context with mute.read or block.read.

### Changelog

- **2026-08-13** · fix · Updated video_total_views definition (`x-api-video-total-views-2026-08-13`, medium)
  X Ads API Analytics VIDEO metric video_total_views now counts views that are 100% in-view for at least 3 seconds, plus any view where a user clicks play. The prior MRC 50%-in-view-for-2s definition no longer applies; video_3s100pct_views and VIEW_3S_100PCT bidding remain.
- **2026-07-21** · feature · Mute and block events for X Activity API (`x-api-mute-block-2026-07-21`, high)
  Subscribe by user id for mute/unmute and block/unblock. Private events requiring user-context OAuth 2.0.
- **2026-06-04** · feature · Post create and delete events (`x-api-post-events-2026-06-04`, high)
  post.create and post.delete on the Activity API. For keyword filtering, continue to use Filtered Stream.
- **2026-03-18** · feature · Direction filter and event-type rename (`x-api-direction-2026-03-18`, high)
  Optional direction filter on Activity API subscriptions. Event type naming moved off PascalCase.

## Playbooks

Composition recipes only. Every product_id and feature_id must already exist in catalog/catalog.json. Do not invent capabilities, versions, or changelog rows. Scheduled scrape automations must not add playbooks.

### Grok Bot: read the live catalog before acting (`grok-bot-operator`)

- when: Any agent session that needs current SpaceXAI, Cursor, or X capability facts — especially a Grok Bot that landed on the human UI.
- confidence: high
- products: `grok-bot`
- features: `grok-bot-computer`, `grok-bot-connectors`, `grok-bot-always-on`

If you are a Grok Bot (or any agent) starting from xstack.grok.me or this JSON host, load the live Origin-published files first. Do not invent product features, versions, or changelog rows. Route the user's task through the matching playbook and the official URLs on each product.

1. **Fetch the agent contract from the JSON host**
   GET https://x-stacks.vercel.app/llms.txt (also at https://x-stacks.vercel.app/.well-known/llms.txt). If you found a /llms.txt on https://xstack.grok.me, treat that host as the human UI. Same-origin /catalog.json must 404 or 301 to this host — do not use a UI seed. Always load JSON from https://x-stacks.vercel.app.
   https://x-stacks.vercel.app/llms.txt
2. **First hop: index + status, then catalog/playbooks as needed**
   GET https://x-stacks.vercel.app/index.json (slim product + playbook router) and https://x-stacks.vercel.app/status.json (scrape health). Then GET catalog.json and playbooks.json. generated_at is last successful scrape; published_at is a deploy. If last_run_failed, or now is after last_run_fresh_until, say the scrape failed or is overdue. If status.json marks a source 403 or skipped, or a product has freshness=stale / blocked_pages, do not reconstruct that page from memory.
   https://x-stacks.vercel.app/index.json
3. **Do not invent**
   Answer only from products[], features[], channels[], changelog[], playbooks[], index.json, status.json, and updates.json. Every changelog row has sources[] with a URL. Confidence high is official changelog/announcement mapping; medium needs a human look. Return the answer contract from llms.txt.
4. **Route by task**
   Pick the playbook whose `when` matches. Grok Bot itself is an always-on teammate with its own cloud computer and connectors — including tools with no clean API or MCP — not a substitute for this catalog.
   https://docs.x.ai/grok-bot/overview

### Grok Bot routine: watch X Stack and brief on changes (`grok-bot-watch-catalog`)

- when: The user wants a Grok Bot to keep watching the catalog and message them when something changes — not a one-shot fetch.
- confidence: medium
- products: `grok-bot`
- features: `grok-bot-routines`, `grok-bot-messaging`, `grok-bot-approvals`, `grok-bot-always-on`

Grok Bot can save a workflow as a routine, message like a colleague, and only pull you in for approvals. Use that to poll X Stack: compare generated_at, read updates.json, brief new changelog ids, and escalate medium-confidence rows.

1. **Remember the last generated_at**
   Store the last seen generated_at from https://x-stacks.vercel.app/index.json (or catalog.json). That field is the scrape clock. Ignore published_at moving alone. Do not scrape vendor HTML for this loop.
   https://x-stacks.vercel.app/index.json
2. **Poll on a routine**
   Grok Bot can follow a job, save the workflow as a routine, and run it on its own after the client is closed. On each run, GET index.json and status.json. If last_run_failed, say the scrape failed or is overdue. If generated_at is unchanged, stop (a UI or flatten publish is not news). If it changed, continue.
   https://docs.x.ai/grok-bot/overview
3. **Read the delta, not the whole catalog**
   GET https://x-stacks.vercel.app/watch.json and pick the user's subscribe slices (imagine, bot, build-alpha, voice, cursor, x-api). Then GET updates.json or /updates/{filter}.json. Each row has breaking, can (the verb, e.g. alpha pointer is 1.0.8), and url. changelog.json has the same fields on every historical row. count 0 means this scrape added no changelog rows — say that; do not invent news. Do not treat the grok.me homepage “this week” HTML as the feed.
   https://x-stacks.vercel.app/watch.json
4. **Check scrape health and product freshness**
   GET status.json and index.json. If last_run_failed, or now is after last_run_fresh_until, cite last_run_error and say overdue. If a source is 403/skipped, or a product has freshness=stale (403 keep-prior is stale, not freshly scraped), say that. Do not fetch x.ai/news or x.ai/build until a docs/RSS URL returns 200. Grok Build versions come from x.ai/cli/*.
   https://x-stacks.vercel.app/status.json
5. **Message a short briefing; approve only medium/low**
   Message like a colleague with the answer contract (playbook_id, product_ids, answer, sources, confidence, missing). High-confidence rows can be a briefing. Medium or low rows: pull the human in for a judgment call (approvals). Do not treat Ads video_total_views as an Activity API feature.
   https://x.ai/news/introducing-grok-bot

### Always-on: Grok Bot vs Cursor Cloud Agents (`always-on-route`)

- when: The user wants something that keeps working after they close the laptop, and you must choose Grok Bot, Cursor Cloud Agents, or say both apply to different parts of the job.
- confidence: medium
- products: `grok-bot`, `cursor-cloud-agents`
- features: `grok-bot-computer`, `grok-bot-always-on`, `grok-bot-connectors`, `grok-bot-routines`, `grok-bot-approvals`, `cursor-ca-subscriptions`, `cursor-ca-goal`, `cursor-ca-subagents`, `cursor-ca-builds`

Both products are always-on agents. The catalog records different jobs: Grok Bot is a teammate with its own cloud computer and connectors (including apps with no clean API or MCP). Cursor Cloud Agents run in Cursor's harness — PR/Slack/schedule subscriptions, /goal, isolated-VM subagents, and Origin repos.

1. **Choose Grok Bot when the work is a teammate on a cloud computer**
   Grok Bot (early beta): each bot has a cloud computer, can sign into apps and websites including tools with no clean API or MCP, messages like a colleague, saves routines, and only pulls you in for approvals. Availability recorded as desktop and iOS for SuperGrok Heavy, Cursor Ultra, and Cursor Teams Premium.
   https://x.ai/news/introducing-grok-bot
2. **Choose Cursor Cloud Agents when the work is repo/event-driven in Cursor**
   Cloud Agents subscribe to PRs, a Slack thread, or a schedule; hold a /goal until it is met (pair with /loop); spawn subagents on isolated VMs; accept steering without interrupting the current turn. Builds pre-warm the environment.
   https://cursor.com/changelog/08-19-26
3. **Do not collapse the two products**
   The catalog does not describe Grok Bot and Cursor Cloud Agents as one system. If the job needs both a signed-in cloud computer and PR/Slack subscriptions, say so and keep the product_ids separate.

### Long-running code work on Grok 4.6 (`long-running-code`)

- when: The user wants a long coding or product-building trajectory — plan, edit, test — using Grok 4.6 rather than a different model.
- confidence: high
- products: `grok-4-6`, `grok-build`, `cursor-desktop`, `grok-api`
- features: `grok-4-6-long-running-agents`, `grok-4-6-self-testing`, `grok-4-6-visual-interactive`, `grok-4-6-availability`, `grok-build-plan-mode`, `grok-build-subagents`, `cursor-desktop-grok-4-6`

Grok 4.6 is documented for staying with multi-step research, analysis, and codebase work, with self-testing on longer trajectories. The catalog records it in Grok Build, Cursor, and the xAI API the same day it shipped.

1. **Use Grok 4.6 for long trajectories**
   Documented focus: stay with complex tasks across many steps — research, analysis, codebase work, and turning an idea into a polished application. 500k context, text and image in, text-only out, no text output limit. On longer trajectories it checks its own work before moving on.
   https://x.ai/news/grok-4-6
2. **Grok Build for plan-then-build in the SpaceXAI coding agent**
   Grok Build is powered by Grok 4.6. Plan mode blocks edits until the plan is approved. Subagents run in parallel with their own context windows and worktrees. Skills, hooks, MCP, and AGENTS.md adapt it to a repo.
   https://docs.x.ai/build/overview
3. **Cursor desktop/CLI as the local control plane**
   Grok 4.6 is available in Cursor (recorded 2026-08-12). Cursor desktop/CLI is documented as the local control plane for Cloud Agents, Origin, and Grok 4.6.
   https://x.ai/news/grok-4-6
4. **API when you need the model without those agents**
   Same flagship via the xAI API / console.x.ai: 500k context, reasoning effort low/medium/high/xhigh. See playbook grok-4-6-surfaces for partner hosts.
   https://docs.x.ai/developers/grok-4-6

### Install Grok Build (stable / alpha / enterprise) (`install-grok-build-channels`)

- when: The user wants to install or switch Grok Build channels.
- confidence: high
- products: `grok-build`
- features: `grok-build-channels`

Grok Build ships three installer channels via GROK_CHANNEL. Read live versions from Product.channels on grok-build (or the x.ai/cli pointer pages). Do not invent a version number.

1. **Read the live channel versions from the catalog**
   Open products[id=grok-build].channels[]. Each row has id, version, updated, install, notes. Official pointers are plain version strings at https://x.ai/cli/stable, https://x.ai/cli/alpha, and https://x.ai/cli/enterprise.
   https://x.ai/cli/stable
2. **Stable (default)**
   curl -fsSL https://x.ai/cli/install.sh | bash
   https://x.ai/cli/stable
3. **Alpha**
   curl -fsSL https://x.ai/cli/install.sh | GROK_CHANNEL=alpha bash. Persist in ~/.grok/config.toml under [cli] channel. The welcome screen shows the active channel.
   https://x.ai/cli/alpha
4. **Enterprise**
   curl -fsSL https://x.ai/cli/install.sh | GROK_CHANNEL=enterprise bash
   https://x.ai/cli/enterprise
5. **Do not invent versions**
   If the catalog and the pointer page disagree, cite both and prefer the pointer body. Feature notes for a new pointer may not have landed on https://x.ai/build/changelog yet.

### Cursor desktop: Latest vs Early Access vs Nightly (`cursor-desktop-tracks`)

- when: The user asks which Cursor desktop build to run, or how to get a pre-release.
- confidence: high
- products: `cursor-desktop`
- features: `cursor-desktop-prerelease`

cursor.com/download is Latest (stable) only. Early Access and Nightly are in-app: Settings → Beta → Update Access. The catalog must not invent a nightly version number; keep version as pre-release / nightly and put the fact in notes.

1. **Latest (stable)**
   https://cursor.com/download — macOS, Windows, and Linux. Read the live version from products[id=cursor-desktop].channels[id=stable]. Feature write-ups on the changelog are per minor, not per patch build.
   https://cursor.com/download
2. **Early Access**
   Settings → Beta → Update Access. Desktop minors often land on Early Access before the default track. Channel version in the catalog is "pre-release" when no public pointer exists.
3. **Nightly**
   Settings → Beta → Update Access → Nightly. Cursor does not publish a separate nightly changelog or public version pointer. Do not invent a nightly number.

### Origin repos with Cursor Cloud Agents (`origin-plus-cloud-agents`)

- when: The user wants hosted Cursor repos plus always-on agents that wake on PRs or a goal — not a Grok Bot cloud-computer teammate.
- confidence: high
- products: `cursor-origin`, `cursor-cloud-agents`, `cursor-desktop`
- features: `cursor-origin-repos`, `cursor-origin-agents`, `cursor-origin-prs`, `cursor-origin-github`, `cursor-ca-subscriptions`, `cursor-ca-goal`, `cursor-ca-builds`, `cursor-desktop-cli`

Cursor desktop/CLI is documented as the local control plane for Cloud Agents and Origin. Origin hosts repos (early beta, paid plans) and records that agents can query code, change files, update PRs, and push branches. Cloud Agents subscribe to PRs, hold a /goal, and boot faster from Builds.

1. **Host the repo on Origin**
   Create and host from the Codebase tab. URL form cursor.com/codebase/{name}. Clone and push with the CLI. GitHub sync is bidirectional and can be disconnected. Early beta on paid plans.
   https://cursor.com/changelog/origin-code-hosting
2. **Agents on hosted repos**
   Origin records that agents can query code, make changes, update PRs, and push branches from every hosted repo. App extensions include Vercel PR previews and Depot/Buildkite CI.
   https://cursor.com/changelog/origin-code-hosting
3. **Cloud Agents for events and goals**
   Subscribe to PRs (agents can auto-subscribe to PRs they create), watch a Slack thread, or run a schedule. Assign /goal for a long-lived objective; pair with /loop. Builds are included with Cloud Agents.
   https://cursor.com/changelog/08-19-26
4. **Drive it from desktop or CLI**
   Cursor IDE/CLI clones and pushes Origin repos and drives agents. That is the documented local control plane — not a second hosting product.
   https://cursor.com/download

### X Activity API: mute/block and post events (`x-activity-api`)

- when: The user wants realtime X user-activity webhooks — mutes, blocks, or post create/delete — not Ads analytics metrics.
- confidence: high
- products: `x-api`
- features: `x-api-activity`, `x-api-mute-block`, `x-api-webhooks-billing`, `x-api-ppu`

X API v2 is pay-per-usage. The X Activity API delivers realtime user-activity webhooks (successor path from the legacy Account Activity API). Documented event additions include mute/block (2026-07-21) and post create/delete (2026-06-04). Inbound webhook events are billed per delivery with 24-hour UTC dedup.

1. **Start at Activity API + pricing**
   Docs: https://docs.x.com/x-api/activity/introduction. Pricing is pay-per-usage credits (no subscription SKUs). Owned reads vs public lookup are priced separately.
   https://docs.x.com/x-api/activity/introduction
2. **Mute and block events**
   mute.mute / mute.unmute / block.block / block.unblock. Private events; OAuth 2.0 user-context with mute.read or block.read. Subscribe by user id.
   https://docs.x.com/changelog
3. **Post create and delete**
   post.create and post.delete on the Activity API. For keyword filtering, the changelog says continue to use Filtered Stream.
   https://docs.x.com/changelog
4. **Billing and filters already in the catalog**
   Inbound Activity API events cost per delivery, same 24h UTC dedup window; many outbound actions do not. Optional direction filter on subscriptions is recorded (2026-03-18). Do not treat Ads API Analytics metrics (including video_total_views) as Activity API features.

### Imagine + Voice on the xAI API (`imagine-voice-api`)

- when: The user wants image/video generation or realtime voice through the xAI API rather than only grok.com UI surfaces.
- confidence: high
- products: `grok-imagine`, `grok-voice`, `grok-api`
- features: `grok-imagine-image-2`, `grok-imagine-video-1-5`, `grok-imagine-api`, `grok-voice-s2s`, `grok-voice-tts`, `grok-voice-stt`, `grok-voice-think-fast-2`, `grok-api-media-voice`, `grok-api-console`

The xAI API and console.x.ai expose Grok models, Imagine, and Voice on one API. Imagine covers Image 2.0 and Video 1.5 (generate and edit). Voice covers speech-to-speech, TTS, and STT (Voice Think Fast 2.0).

1. **One API / console**
   Keys, playground, usage: https://console.x.ai. Media and voice are recorded as the same API as flagship text models.
   https://console.x.ai
2. **Imagine images and video**
   Image 2.0 Quality Mode on grok.com/imagine and apps; API id grok-imagine-image-2.0. Video 1.5: text-to-video, image-to-video, reference-to-video; native 1080p for T2V/I2V. Editing: magic wand, segmentation, background removal, up to 5 reference images. Image input billed separately from output.
   https://docs.x.ai/developers/model-capabilities/imagine
3. **Voice realtime, TTS, STT**
   Speech-to-speech starts at $0.05 / min; TTS is $15.00 / 1M characters; STT has adjustable vad_threshold (0 disables the gate). Current Voice Think Fast generation recorded as 2.0 (2026-07-29).
   https://docs.x.ai/developers/model-capabilities

### Grok 4.6 via API, console, and partners (`grok-4-6-surfaces`)

- when: The user asks where to call Grok 4.6, how it is priced on the xAI API, or which partner hosts are documented.
- confidence: high
- products: `grok-4-6`, `grok-api`
- features: `grok-4-6-availability`, `grok-4-6-pricing`, `grok-4-6-reasoning-effort`, `grok-4-6-context`, `grok-api-models`, `grok-api-console`, `grok-api-partners`

The catalog records Grok 4.6 in Cursor, Grok Build, the xAI API, and partners: OpenRouter, Vercel, Cloudflare, Amazon Bedrock, and GitHub Copilot. Use this when the question is where the model runs, not how to operate Grok Build or Cursor agents.

1. **First-party API and console**
   https://console.x.ai for keys and playground. Docs: https://docs.x.ai/developers/grok-4-6 and release notes. 500k context; reasoning low / medium / high (default) / xhigh.
   https://console.x.ai
2. **Documented API pricing**
   $2 / $0.50 / $6 per 1M tokens (input / cached / output) below 200k prompt tokens; $4 / $1 / $12 above. Fast variant is twice the price. Confirm on the official docs if you need current numbers.
   https://docs.x.ai/developers/release-notes
3. **Documented hosts**
   Cursor, Grok Build, xAI API, OpenRouter, Vercel, Cloudflare, Amazon Bedrock, GitHub Copilot. Availability changelog rows exist for Bedrock (2026-08-19) and GitHub Copilot (2026-08-14).
   https://x.ai/news/grok-4-6