# Codex — built-in provider manifest.
# Authoring format for the def half. See packages/hive-agents/src/manifest.ts.
manifestVersion: 2
# Written for the app that hands codex its trust for our hooks (hooks.trust).
minAppVersion: '2026.10.2'
id: codex
label: Codex
bin: codex
enabled: true
caps:
  promptDelivery: typed
  turnSignal: true
  resume: cwd
  supervise: human
  blockedDetection: true
icon:
  viewBox: 0 0 24 24
  attrs:
    fill: currentColor
    fillRule: evenodd
  shapes:
  - path:
      d: M8.086.457a6.105 6.105 0 013.046-.415c1.333.153 2.521.72 3.564 1.7a.117.117 0 00.107.029c1.408-.346
        2.762-.224 4.061.366l.063.03.154.076c1.357.703 2.33 1.77 2.918 3.198.278.679.418 1.388.421 2.126a5.655
        5.655 0 01-.18 1.631.167.167 0 00.04.155 5.982 5.982 0 011.578 2.891c.385 1.901-.01 3.615-1.183
        5.14l-.182.22a6.063 6.063 0 01-2.934 1.851.162.162 0 00-.108.102c-.255.736-.511 1.364-.987 1.992-1.199
        1.582-2.962 2.462-4.948 2.451-1.583-.008-2.986-.587-4.21-1.736a.145.145 0 00-.14-.032c-.518.167-1.04.191-1.604.185a5.924
        5.924 0 01-2.595-.622 6.058 6.058 0 01-2.146-1.781c-.203-.269-.404-.522-.551-.821a7.74 7.74 0
        01-.495-1.283 6.11 6.11 0 01-.017-3.064.166.166 0 00.008-.074.115.115 0 00-.037-.064 5.958 5.958
        0 01-1.38-2.202 5.196 5.196 0 01-.333-1.589 6.915 6.915 0 01.188-2.132c.45-1.484 1.309-2.648 2.577-3.493.282-.188.55-.334.802-.438.286-.12.573-.22.861-.304a.129.129
        0 00.087-.087A6.016 6.016 0 015.635 2.31C6.315 1.464 7.132.846 8.086.457zm-.804 7.85a.848.848
        0 00-1.473.842l1.694 2.965-1.688 2.848a.849.849 0 001.46.864l1.94-3.272a.849.849 0 00.007-.854l-1.94-3.393zm5.446
        6.24a.849.849 0 000 1.695h4.848a.849.849 0 000-1.696h-4.848z
# A message typed while it works is read after its next tool call, in that turn: codex 0.163.0
# shows "Messages to be submitted after next tool call".
steer: true
# An image whose path is pasted alone is attached: codex 0.163.0 shows "[Image #1]".
pastedImages: true
options:
- id: model
  label: Model
  flag: --model
- id: mode
  label: Approval
  flag: --ask-for-approval
  default: on-request
  values:
    yolo:
    - --dangerously-bypass-approvals-and-sandbox
- id: sandbox
  label: Sandbox
  flag: --sandbox
  default: workspace-write
install:
  url: https://developers.openai.com/codex/cli
  command: "npm install -g @openai/codex"
detect:
  # Only the live prompt area says what the agent is waiting for: its own streamed output
  # quotes files and plans full of "do you want … yes", and a screen-wide match reads those
  # as a question and notifies for each one.
  scope:
    kind: tail
    n: 20
  rules:
  # What it is WAITING on is its own chooser or its own footer, on one line. Matching the
  # words a question is made of ("do you want", "[y/n]", "yes") anywhere on the screen read
  # the output it was streaming back as a question, over and over.
  - when:
      any:
      - seq:
        - any: ["\u203a", ">"]
        - run: space
          min: 1
        - run: digit
          min: 1
        - lit: '.'
        - run: space
          min: 1
        at: start
      - line:
        - contains: press enter to confirm or esc to cancel
      - line:
        - contains: enter to submit answer
      - line:
        - contains: allow command?
      # Its hook review, which a first tile in a new profile always lands on.
      - line:
        - contains: hook needs review
      - line:
        - contains: hooks need review
    then: blocked
  - when:
      helper: hasInterruptPattern
    then: working
  - when:
      line:
      - startsWith: •
      - containsCS: Working (
    then: working
  default: idle
session:
  resume:
    args:
    - resume
    - '{id}'
    find:
      strategy: jsonl-header
      root: '{home}/.codex/sessions'
      cwdPath: payload.cwd
      idPath: payload.id
      require:
        type: session_meta
# Codex runs a hook only once its user has trusted that exact command. The commands are the same
# in every tile (the tile is read from the spawn environment), and the document carries codex's
# trust for these hooks and no others, so it never asks about ours and still asks about any other.
hooks:
  arg: -c
  format: toml
  stable: true
  trust: codex
  template: 'hooks={events}'
  events:
    SessionStart:
      emit: session.started
      timeout: 10
    SessionEnd:
      emit: session.ended
      timeout: 3
    UserPromptSubmit:
      emit: turn.started
      timeout: 10
    Stop:
      hook: turnEnd
      timeout: 10
    Interrupt:
      emit: turn.ended
      outcome: interrupted
      timeout: 3
    PermissionRequest:
      emit: input.requested
      kind: permission
      timeout: 10
    PostToolUse:
      emit: input.resolved
      timeout: 10
    SubagentStart:
      emit: subagent.started
      timeout: 10
    SubagentStop:
      emit: subagent.stopped
      timeout: 10
    PreCompact:
      emit: compacting.started
      timeout: 10
    PostCompact:
      emit: compacting.ended
      timeout: 10
assets:
- name: hive-turn-end.cjs
  file: hive-turn-end.cjs
  hook: turnEnd
  produces: [turn.ended]
# A tile Hivemind starts must not open on Codex's own update prompt, where a message sent to the
# tile would answer it; the user updates Codex from a terminal instead.
launch:
  hcp: true
  args:
  - -c
  - check_for_update_on_startup=false
  # Its fullscreen transcript takes the wheel for itself and then does nothing with it, so a
  # tile cannot be scrolled at all. In scrollback mode the terminal keeps the wheel and the
  # transcript lands in the tile's own scrollback. `/tui` switches it back.
  - -c
  - tui.fullscreen_transcript=false
commands:
  builtIn:
    - name: "permissions"
      description: "Set what Codex can do without asking first"
    - name: "ide"
      description: "Include open files, current selection, and other IDE context"
    - name: "keymap"
      description: "Remap TUI keyboard shortcuts"
    - name: "vim"
      description: "Toggle Vim mode for the composer"
    - name: "setup-default-sandbox"
      description: "Set up the elevated agent sandbox (Windows only)"
    - name: "agents"
      description: "Open the agent command center"
    - name: "subagents"
      description: "Switch the active subagent thread"
    - name: "apps"
      description: "Browse apps (connectors) and insert them into your prompt"
    - name: "plugins"
      description: "Browse installed and discoverable plugins"
    - name: "hooks"
      description: "View and manage lifecycle hooks"
    - name: "clear"
      description: "Clear the terminal and start a fresh chat"
    - name: "rename"
      description: "Rename the current chat"
    - name: "archive"
      description: "Archive the current session"
    - name: "delete"
      description: "Permanently delete the current session"
    - name: "compact"
      description: "Summarize the visible chat to free tokens"
    - name: "copy"
      description: "Copy the latest completed Codex output"
    - name: "diff"
      description: "Show the Git diff, including files Git isn\u2019t tracking yet"
    - name: "exit"
      description: "Exit the CLI (same as /quit)"
    - name: "experimental"
      description: "Toggle experimental features"
    - name: "approve"
      description: "Approve one retry of a recent auto review denial"
    - name: "memories"
      description: "Configure memory use and generation"
    - name: "skills"
      description: "Browse and use skills"
    - name: "import"
      description: "Import Claude Code or Cursor setup, projects, and chats"
    - name: "feedback"
      description: "Send logs to the Codex maintainers"
    - name: "init"
      description: "Generate an AGENTS.md scaffold in the current directory"
    - name: "logout"
      description: "Sign out of Codex"
    - name: "mcp"
      description: "List configured Model Context Protocol (MCP) tools"
    - name: "mention"
      description: "Attach a file to the chat"
    - name: "model"
      description: "Choose the active model (and reasoning effort, when available)"
    - name: "fast"
      description: "Toggle a Fast service tier when the model catalog exposes one"
    - name: "plan"
      description: "Switch to plan mode and optionally send a prompt"
    - name: "goal"
      description: "Set, edit, pause, resume, view, or clear a task goal"
    - name: "ps"
      description: "Show background terminals and their recent output"
    - name: "stop"
      description: "Stop all background terminals"
    - name: "fork"
      description: "Fork the current chat into a new chat"
    - name: "side"
      description: "Start an ephemeral side chat"
    - name: "raw"
      description: "Toggle raw scrollback mode"
    - name: "resume"
      description: "Resume a saved chat from your session list"
    - name: "new"
      description: "Start a new chat inside the same CLI session"
    - name: "quit"
      description: "Exit the CLI"
    - name: "review"
      description: "Ask Codex to review your working tree"
    - name: "status"
      description: "Display session configuration and token usage"
    - name: "usage"
      description: "View account token usage or use a rate-limit reset"
    - name: "debug-config"
      description: "Print config layer and requirements diagnostics"
    - name: "statusline"
      description: "Configure TUI status-line fields interactively"
    - name: "title"
      description: "Configure terminal window or tab title fields interactively"
    - name: "theme"
      description: "Choose a syntax-highlighting theme"
    - name: "pets"
      description: "Choose or hide a terminal pet"
    - name: "tui"
      description: "Choose the terminal mode for the next launch"
    - name: "export"
      description: "Export the conversation as Markdown"
    - name: "recap"
      description: "Generate a summary of the current conversation"
    - name: "cd"
      description: "Change the session\u2019s working directory"
    - name: "pwd"
      description: "Show the working directory"
    - name: "worktree"
      description: "Start or continue a chat in a managed Git worktree"
    - name: "daemon"
      description: "Open background server controls"
    - name: "warnings"
      description: "View retained warnings and diagnostics"
  folders:
    - path: "{home}/.codex/prompts"
      prefix: "prompts:"