CLI Reference

FLOW Job orchestration and workflows

USAGE flow [command]

COMMANDS action Execute or list workspace actions from a recipe add Add a new job to an existing plan chat Start or manage a chat-based job from a file complete Mark a job as completed completion Generate the autocompletion script for the specified shell config Get or set plan configuration values current Show the current active job plan directory docs Print the structured JSON documentation for this tool finish Finish and clean up a plan and its associated worktree graph Visualize job dependency graph hold Set a plan’s status to ‘hold’ job Manage jobs in orchestration plans list List all plans in the configured plans directory or across all workspaces models List available LLM models for use in jobs and chats open Open a plan’s worktree in a dedicated tmux session plan Manage multi-step orchestration plans resume Resume a completed interactive agent job review Mark a plan as ready for review and execute completion hooks run Run jobs in an orchestration plan set Set the active job plan directory starship Manage Starship prompt integration status Show plan status in an interactive TUI step Step through plan execution interactively tmux Tmux window management commands unhold Remove a plan’s on-hold status unset Clear the active job plan directory version Print the version information for this binary

Flags: -c/—config, -h/—help, —json, -v/—verbose

Use “flow [command] —help” for more information.

FLOW ACTION Execute or list workspace actions from a recipe

Execute a named workspace action defined in the recipe’s workspace_init.yml, or list all available actions.

USAGE flow action [action-name] [plan-name] [flags]

FLAGS -h, —help help for action —list List available actions for the plan

EXAMPLES # List available actions for the current plan flow action —list

# List available actions for a specific plan flow action —list my-plan

# Run the “start-dev” action for the current plan flow action start-dev

# Run the “start-dev” action for a specific plan flow action start-dev my-plan

# Run the init actions (special case) flow action init

Available actions are defined in the recipe’s workspace_init.yml under the ‘actions’ key. The special action ‘init’ runs the actions defined under the ‘init’ key.

FLOW ADD Add a new job to an existing plan

Add a new job to an existing orchestration plan. Can be used interactively or with command-line arguments. If no directory is specified, uses the active job if set.

USAGE flow add [directory] [flags]

FLAGS -d, —depends-on Dependencies (job filenames) —git-changes Include git changes (staged and unstaged) as context for this job -h, —help help for add —include Comma-separated list of files to include as context —inline File types to inline in prompt: • dependencies - Embed dependency output directly in prompt (vs separate files) • include - Embed —include file content directly in prompt • context - Embed project context (.grove/context) in prompt • all - Inline all of the above • files - Inline dependencies + include (not context) • none - No inlining; content provided as separate files (default) -i, —interactive Interactive mode -m, —model LLM model to use for this job (e.g., gemini-3-pro-preview, claude-sonnet-4-20250514) —prepend-dependencies [DEPRECATED] Use —inline=dependencies. Inline dependency content into prompt body -p, —prompt Inline prompt text (alternative to —prompt-file) -f, —prompt-file File containing the prompt —recipe Name of a recipe to add to the plan —recipe-vars Variables for the recipe templates (e.g., key=value) —rules-file Path to a custom rules file for this job —source-file Origin file path for tracking job provenance (e.g., Claude plan file) —template Name of the job template to use —title Job title -t, —type Job type: (default: interactive_agent) • oneshot - Single LLM call, no tools or iteration • chat - Interactive conversation requiring user input • shell - Execute shell commands directly • headless_agent - Autonomous agent without user interaction • interactive_agent - Agent with user interaction (default) • file - Static file content, no execution —worktree Explicitly set the worktree name (overrides automatic inference)

EXAMPLES # Add a job with inline prompt flow add myplan -t agent —title “implement-feature” -d 01-plan.md -p “Implement the feature”

# Add a job with prompt from file flow add myplan -t agent —title “implement-feature” -d 01-plan.md -f prompt.md

# Add a job with prompt from stdin echo “Implement feature X” | flow add myplan -t agent —title “implement-feature” -d 01-plan.md

# Use active job (after: flow set myplan) flow add -t agent —title “implement-feature” -d 01-plan.md -p “Implement feature”

# Specify a model for this job flow add myplan —title “analyze-codebase” —model gemini-3-pro-preview -p “Analyze the codebase”

# Include git changes as context (useful for review/commit jobs) flow add myplan —title “review-changes” —git-changes -p “Review uncommitted changes”

# Use a custom rules file flow add myplan —title “update-docs” —rules-file docs/.rules -p “Update documentation”

# Use a job template flow add myplan —template code-review —title “code-review” -d 01-implement.md

# Include specific files as context flow add myplan —title “refactor-api” —include src/api.go,src/types.go -p “Refactor the API”

# Inline dependency output into the prompt flow add myplan —title “fix-issues” -d 01-review.md —inline=dependencies -p “Fix the issues”

# Add jobs from a recipe flow add myplan —recipe standard-feature —recipe-vars “feature=auth”

FLOW CHAT Start or manage a chat-based job from a file

Initializes a markdown file as a runnable chat job by adding the necessary frontmatter.

USAGE flow chat [flags] flow chat [command]

COMMANDS list List all chat jobs in the configured chat directory run Run outstanding chat jobs that are waiting for an LLM response

Flags: -h/—help, -m/—model, -s/—spec-file, -t/—title

EXAMPLES flow chat -s /path/to/my-notes/new-feature.md

Use “flow chat [command] —help” for more information.

FLOW CHAT LIST List all chat jobs in the configured chat directory

USAGE flow chat list [flags]

FLAGS -h, —help help for list —status Filter chats by status (e.g., pending_user, completed, running)

FLOW CHAT RUN Run outstanding chat jobs that are waiting for an LLM response

Scans the configured chat directory for chats where the last turn is from a user and executes them sequentially to generate the next LLM response.

You can optionally specify chat titles to run only specific chats: flow chat run # Run all pending chats flow chat run testing-situation # Run only the chat titled “testing-situation” flow chat run chat1 chat2 # Run multiple specific chats

USAGE flow chat run [title…] [flags]

FLAGS -h, —help help for run

FLOW COMPLETE Mark a job as completed

Mark a job as completed. This is especially useful for chat jobs that would otherwise remain in pending_user status indefinitely.

USAGE flow complete <job-file> [flags]

FLAGS -h, —help help for complete

EXAMPLES # Complete a chat job flow complete my-project/plan.md

FLOW CONFIG Get or set plan configuration values

Get or set configuration values in the plan’s .grove-plan.yml file.

USAGE flow config [directory] [flags]

FLAGS —get Get a configuration value -h, —help help for config —json Output in JSON format —set Set a configuration value (format: key=value)

EXAMPLES # Set a single value flow plan config myplan —set model=gemini-2.0-flash

# Set multiple values flow plan config myplan —set model=gemini-2.0-flash —set worktree=feature/new

# Get a value flow plan config myplan —get model

# Show all configuration flow plan config myplan

FLOW CURRENT Show the current active job plan directory

Show the current active job plan directory.

If no active job is set, this command will indicate that.

USAGE flow current [flags]

FLAGS -h, —help help for current

FLOW DOCS Print the structured JSON documentation for this tool

This command outputs the structured documentation for this tool in JSON format, which is used by other ecosystem tools like grove-mcp.

USAGE flow docs [flags]

FLAGS -h, —help help for docs

FLOW FINISH Finish and clean up a plan and its associated worktree

Guides through the process of cleaning up a completed plan. This can include removing the git worktree, deleting the branch, closing tmux sessions, and archiving the plan.

USAGE flow finish [directory] [flags]

FLAGS —archive Archive the plan directory to a local .archive subdirectory —clean-dev-links Clean up development binary links from the worktree —close-session Close the associated tmux session —delete-branch Delete the local git branch —delete-remote Delete the remote git branch —force Force git operations (use with caution) -h, —help help for finish —prune-worktree Remove the git worktree directory —rebuild-binaries Rebuild binaries in the main repository -y, —yes Automatically confirm all cleanup actions

FLOW GRAPH Visualize job dependency graph

Generate a visualization of the job dependency graph. Supports multiple output formats including Mermaid, DOT, and ASCII. If no directory is specified, uses the active job if set.

USAGE flow graph [directory] [flags]

FLAGS -f, —format Output format: (default: mermaid) • mermaid • dot • ascii -h, —help help for graph -o, —output Output file (stdout if not specified) -p, —port Port for web server (default: 8080) -s, —serve Serve interactive HTML visualization

FLOW HOLD Set a plan's status to 'hold'

Sets the status of a plan to ‘hold’ in its .grove-plan.yml file. On-hold plans are hidden from most views by default to reduce clutter.

USAGE flow hold [directory] [flags]

FLAGS -h, —help help for hold

FLOW JOB Manage jobs in orchestration plans

Commands for managing jobs in orchestration plans.

USAGE flow job [command]

COMMANDS add Add a new job to an existing plan

Flags: -h/—help

Use “flow job [command] —help” for more information.

FLOW JOB ADD Add a new job to an existing plan

Add a new job to an existing orchestration plan. Can be used interactively or with command-line arguments. If no directory is specified, uses the active job if set.

USAGE flow job add [directory] [flags]

FLAGS -d, —depends-on Dependencies (job filenames) —git-changes Include git changes (staged and unstaged) as context for this job -h, —help help for add —include Comma-separated list of files to include as context —inline File types to inline in prompt: • dependencies - Embed dependency output directly in prompt (vs separate files) • include - Embed —include file content directly in prompt • context - Embed project context (.grove/context) in prompt • all - Inline all of the above • files - Inline dependencies + include (not context) • none - No inlining; content provided as separate files (default) -i, —interactive Interactive mode -m, —model LLM model to use for this job (e.g., gemini-3-pro-preview, claude-sonnet-4-20250514) —prepend-dependencies [DEPRECATED] Use —inline=dependencies. Inline dependency content into prompt body -p, —prompt Inline prompt text (alternative to —prompt-file) -f, —prompt-file File containing the prompt —recipe Name of a recipe to add to the plan —recipe-vars Variables for the recipe templates (e.g., key=value) —rules-file Path to a custom rules file for this job —source-file Origin file path for tracking job provenance (e.g., Claude plan file) —template Name of the job template to use —title Job title -t, —type Job type: (default: interactive_agent) • oneshot - Single LLM call, no tools or iteration • chat - Interactive conversation requiring user input • shell - Execute shell commands directly • headless_agent - Autonomous agent without user interaction • interactive_agent - Agent with user interaction (default) • file - Static file content, no execution —worktree Explicitly set the worktree name (overrides automatic inference)

EXAMPLES # Add a job with inline prompt flow job add myplan -t agent —title “implement-feature” -d 01-plan.md -p “Implement the feature”

# Add a job with prompt from file flow job add myplan -t agent —title “implement-feature” -d 01-plan.md -f prompt.md

# Add a job with prompt from stdin echo “Implement feature X” | flow job add myplan -t agent —title “implement-feature” -d 01-plan.md

# Use active job (after: flow set myplan) flow job add -t agent —title “implement-feature” -d 01-plan.md -p “Implement feature”

# Specify a model for this job flow job add myplan —title “analyze-codebase” —model gemini-3-pro-preview -p “Analyze the codebase”

# Include git changes as context (useful for review/commit jobs) flow job add myplan —title “review-changes” —git-changes -p “Review uncommitted changes”

# Use a custom rules file flow job add myplan —title “update-docs” —rules-file docs/.rules -p “Update documentation”

# Use a job template flow job add myplan —template code-review —title “code-review” -d 01-implement.md

# Include specific files as context flow job add myplan —title “refactor-api” —include src/api.go,src/types.go -p “Refactor the API”

# Inline dependency output into the prompt flow job add myplan —title “fix-issues” -d 01-review.md —inline=dependencies -p “Fix the issues”

# Add jobs from a recipe flow job add myplan —recipe standard-feature —recipe-vars “feature=auth”

FLOW LIST List all plans in the configured plans directory or across all workspaces

Scans for and lists orchestration plans. By default, it scans the directory specified in the notebooks configuration. With —all-workspaces, it discovers all projects and scans for plans within them.

USAGE flow list [flags]

FLAGS —all-workspaces List plans across all discovered workspaces -h, —help help for list —include-finished Include finished plans in the output —show-hold Include on-hold plans in the output -v, —verbose Show detailed information including jobs in each plan

FLOW MODELS List available LLM models for use in jobs and chats

Lists recommended LLM models that can be used in job and chat frontmatter.

While other models supported by the ‘llm’ tool may work, these are the primary models tested with Grove Flow.

USAGE flow models [flags]

FLAGS -h, —help help for models

FLOW OPEN Open a plan's worktree in a dedicated tmux session

Switches to or creates a tmux session for the plan’s worktree and opens the interactive status TUI. This provides a one-command entry point into a plan’s interactive environment.

USAGE flow open [directory] [flags]

FLAGS -h, —help help for open

FLOW PLAN Manage multi-step orchestration plans

Manage multi-step orchestration plans in dedicated directories.

USAGE flow plan [command]

COMMANDS action Execute or list workspace actions from a recipe (use: flow action) add Add a new job to an existing plan (use: flow add) complete Mark a job as completed (use: flow complete) config Get or set plan configuration values context Manage job-specific context rules current Show the current active job plan directory extract Extract chat blocks into a new chat job or list available blocks finish Finish and clean up a plan and its associated worktree (use: flow finish) graph Visualize job dependency graph (use: flow graph) hold Set a plan’s status to ‘hold’ init Initialize a new plan directory, interactively or via flags jobs Manage job types and definitions list List all plans (use: flow list) merge-worktree Merge plan worktree branch to main open Open a plan’s worktree in a dedicated tmux session (use: flow open) recipes Manage plan recipes resume Resume a completed interactive agent job review Mark a plan as ready for review and execute completion hooks (use: flow review) run Run jobs (use: flow run) set Set the active job plan directory status Show plan status in an interactive TUI (use: flow status) step Step through plan execution interactively (use: flow step) templates Manage job templates tui Launch interactive TUI for browsing and managing plans unhold Remove a plan’s on-hold status unset Clear the active job plan directory update-worktree Update plan worktree by rebasing on main

Flags: -h/—help

Use “flow plan [command] —help” for more information.

FLOW PLAN ACTION Execute or list workspace actions from a recipe (use: flow action)

Execute a named workspace action defined in the recipe’s workspace_init.yml, or list all available actions.

USAGE flow plan action [action-name] [plan-name] [flags]

FLAGS -h, —help help for action —list List available actions for the plan

EXAMPLES # List available actions for the current plan flow plan action —list

# List available actions for a specific plan flow plan action —list my-plan

# Run the “start-dev” action for the current plan flow plan action start-dev

# Run the “start-dev” action for a specific plan flow plan action start-dev my-plan

# Run the init actions (special case) flow plan action init

Available actions are defined in the recipe’s workspace_init.yml under the ‘actions’ key. The special action ‘init’ runs the actions defined under the ‘init’ key.

FLOW PLAN ADD Add a new job to an existing plan (use: flow add)

Add a new job to an existing orchestration plan. Can be used interactively or with command-line arguments. If no directory is specified, uses the active job if set.

USAGE flow plan add [directory] [flags]

FLAGS -d, —depends-on Dependencies (job filenames) —git-changes Include git changes (staged and unstaged) as context for this job -h, —help help for add —include Comma-separated list of files to include as context —inline File types to inline in prompt: • dependencies - Embed dependency output directly in prompt (vs separate files) • include - Embed —include file content directly in prompt • context - Embed project context (.grove/context) in prompt • all - Inline all of the above • files - Inline dependencies + include (not context) • none - No inlining; content provided as separate files (default) -i, —interactive Interactive mode -m, —model LLM model to use for this job (e.g., gemini-3-pro-preview, claude-sonnet-4-20250514) —prepend-dependencies [DEPRECATED] Use —inline=dependencies. Inline dependency content into prompt body -p, —prompt Inline prompt text (alternative to —prompt-file) -f, —prompt-file File containing the prompt —recipe Name of a recipe to add to the plan —recipe-vars Variables for the recipe templates (e.g., key=value) —rules-file Path to a custom rules file for this job —source-file Origin file path for tracking job provenance (e.g., Claude plan file) —template Name of the job template to use —title Job title -t, —type Job type: (default: interactive_agent) • oneshot - Single LLM call, no tools or iteration • chat - Interactive conversation requiring user input • shell - Execute shell commands directly • headless_agent - Autonomous agent without user interaction • interactive_agent - Agent with user interaction (default) • file - Static file content, no execution —worktree Explicitly set the worktree name (overrides automatic inference)

EXAMPLES # Add a job with inline prompt flow plan add myplan -t agent —title “implement-feature” -d 01-plan.md -p “Implement the feature”

# Add a job with prompt from file flow plan add myplan -t agent —title “implement-feature” -d 01-plan.md -f prompt.md

# Add a job with prompt from stdin echo “Implement feature X” | flow plan add myplan -t agent —title “implement-feature” -d 01-plan.md

# Use active job (after: flow set myplan) flow plan add -t agent —title “implement-feature” -d 01-plan.md -p “Implement feature”

# Specify a model for this job flow plan add myplan —title “analyze-codebase” —model gemini-3-pro-preview -p “Analyze the codebase”

# Include git changes as context (useful for review/commit jobs) flow plan add myplan —title “review-changes” —git-changes -p “Review uncommitted changes”

# Use a custom rules file flow plan add myplan —title “update-docs” —rules-file docs/.rules -p “Update documentation”

# Use a job template flow plan add myplan —template code-review —title “code-review” -d 01-implement.md

# Include specific files as context flow plan add myplan —title “refactor-api” —include src/api.go,src/types.go -p “Refactor the API”

# Inline dependency output into the prompt flow plan add myplan —title “fix-issues” -d 01-review.md —inline=dependencies -p “Fix the issues”

# Add jobs from a recipe flow plan add myplan —recipe standard-feature —recipe-vars “feature=auth”

FLOW PLAN COMPLETE Mark a job as completed (use: flow complete)

Mark a job as completed. This is especially useful for chat jobs that would otherwise remain in pending_user status indefinitely.

USAGE flow plan complete <job-file> [flags]

FLAGS -h, —help help for complete

EXAMPLES # Complete a chat job flow plan complete my-project/plan.md

# Complete any job by its filename flow plan complete my-project/01-design-api.md

FLOW PLAN CONFIG Get or set plan configuration values

Get or set configuration values in the plan’s .grove-plan.yml file.

USAGE flow plan config [directory] [flags]

FLAGS —get Get a configuration value -h, —help help for config —json Output in JSON format —set Set a configuration value (format: key=value)

EXAMPLES # Set a single value flow plan config myplan —set model=gemini-2.0-flash

# Set multiple values flow plan config myplan —set model=gemini-2.0-flash —set worktree=feature/new

# Get a value flow plan config myplan —get model

# Show all configuration flow plan config myplan

FLOW PLAN CONTEXT Manage job-specific context rules

Manage job-specific context rules for plan jobs

USAGE flow plan context [command]

COMMANDS set Save current context rules for a job

Flags: -h/—help

Use “flow plan context [command] —help” for more information.

FLOW PLAN CONTEXT SET Save current context rules for a job

Save the current active .grove/rules file as job-specific context rules.

The rules will be saved in the plan’s rules/ subdirectory and referenced in the job’s frontmatter. When the job runs, it will automatically use these saved rules to generate context.

USAGE flow plan context set <job-file> [flags]

FLAGS -h, —help help for set

FLOW PLAN CURRENT Show the current active job plan directory

Show the current active job plan directory.

If no active job is set, this command will indicate that.

USAGE flow plan current [flags]

FLAGS -h, —help help for current

FLOW PLAN EXTRACT Extract chat blocks into a new chat job or list available blocks

Extract specific LLM responses from a chat into a new chat job for further refinement, or list available blocks in a file.

The special argument “all” extracts all content below the frontmatter. The special argument “list” shows all available block IDs in the file.

USAGE flow plan extract <block-id-1> [block-id-2…] | all | list [flags]

FLAGS -d, —depends-on Dependencies (job filenames) —file Chat file to extract from (default: plan.md) (default: plan.md) -h, —help help for extract —json Output in JSON format (for list command) —model LLM model to use for this job —output Output type: (default: file) • file • commit • none —title Title for the new chat job (required for extract) —worktree Explicitly set the worktree name (overrides automatic inference)

EXAMPLES flow plan extract —title “Database Schema Refinement” f3b9a2 a1c2d4 flow plan extract —file chat-session.md —title “API Design” d4e5f6 flow plan extract all —file doc.md —title “Full Document” flow plan extract list —file chat-session.md flow plan extract list —file chat-session.md —json

FLOW PLAN FINISH Finish and clean up a plan and its associated worktree (use: flow finish)

Guides through the process of cleaning up a completed plan. This can include removing the git worktree, deleting the branch, closing tmux sessions, and archiving the plan.

USAGE flow plan finish [directory] [flags]

FLAGS —archive Archive the plan directory to a local .archive subdirectory —clean-dev-links Clean up development binary links from the worktree —close-session Close the associated tmux session —delete-branch Delete the local git branch —delete-remote Delete the remote git branch —force Force git operations (use with caution) -h, —help help for finish —prune-worktree Remove the git worktree directory —rebuild-binaries Rebuild binaries in the main repository -y, —yes Automatically confirm all cleanup actions

FLOW PLAN GRAPH Visualize job dependency graph (use: flow graph)

Generate a visualization of the job dependency graph. Supports multiple output formats including Mermaid, DOT, and ASCII. If no directory is specified, uses the active job if set.

USAGE flow plan graph [directory] [flags]

FLAGS -f, —format Output format: (default: mermaid) • mermaid • dot • ascii -h, —help help for graph -o, —output Output file (stdout if not specified) -p, —port Port for web server (default: 8080) -s, —serve Serve interactive HTML visualization

FLOW PLAN HOLD Set a plan's status to 'hold'

Sets the status of a plan to ‘hold’ in its .grove-plan.yml file. On-hold plans are hidden from most views by default to reduce clutter.

USAGE flow plan hold [directory] [flags]

FLAGS -h, —help help for hold

FLOW PLAN INIT Initialize a new plan directory, interactively or via flags

Initialize a new orchestration plan in the specified directory. Creates a .grove-plan.yml file with default configuration options. If no directory is provided, an interactive TUI will be launched.

USAGE flow plan init [directory] [flags]

FLAGS —extract-all-from Path to a markdown file to extract all content from into an initial job -f, —force Overwrite existing directory —from-note Path to a note file whose body will be used as the prompt for the first job -h, —help help for init —init Execute init actions from the recipe’s workspace_init.yml —model Default model for jobs (e.g., claude-3-5-sonnet-20241022, gpt-4) —note-ref Path to the source note to link to this plan —note-target-file Filename of the job within the recipe to apply the —from-note content and reference to —open-session Immediately open a tmux session for the plan (uses worktree if configured, otherwise main repo) —recipe Name of a plan recipe to initialize from (e.g., standard-feature). When using —recipe-cmd, this can be omitted if the command provides only one recipe —recipe-cmd Command that outputs JSON recipe definitions (overrides grove.yml’s get_recipe_cmd) —recipe-vars Variables to pass to recipe templates. Can be used multiple times or comma-delimited (e.g., —recipe-vars model=gpt-4 —recipe-vars rules_file=docs.rules OR —recipe-vars “model=gpt-4,rules_file=docs.rules,output_dir=docs”) —repos Specific repos to include in ecosystem worktree (e.g., —repos grove-core,grove-flow). If not specified, all submodules are included -t, —tui Launch interactive TUI to create a new plan —worktree Set default worktree (uses plan name if no value provided)

FLOW PLAN JOBS Manage job types and definitions

Manage job types and definitions for orchestration plans.

USAGE flow plan jobs [command]

COMMANDS list List available job types rename Rename a job and update all dependencies update-deps Update a job’s dependencies

Flags: -h/—help

Use “flow plan jobs [command] —help” for more information.

FLOW PLAN JOBS LIST List available job types

Display all available job types for orchestration plans.

USAGE flow plan jobs list [flags]

FLAGS -h, —help help for list

FLOW PLAN JOBS RENAME Rename a job and update all dependencies

Rename a job file and its title in the frontmatter. This command will also update any dependent jobs to reference the new filename.

USAGE flow plan jobs rename <job-file> <new-title> [flags]

FLAGS -h, —help help for rename

EXAMPLES # Rename a job in the current plan flow plan jobs rename 01-design-api.md “Design REST API”

# Rename a job in a specific plan flow plan jobs rename my-project/01-design-api.md “Design REST API”

FLOW PLAN LIST List all plans (use: flow list)

Scans for and lists orchestration plans. By default, it scans the directory specified in the notebooks configuration. With —all-workspaces, it discovers all projects and scans for plans within them.

USAGE flow plan list [flags]

FLAGS —all-workspaces List plans across all discovered workspaces -h, —help help for list —include-finished Include finished plans in the output —show-hold Include on-hold plans in the output -v, —verbose Show detailed information including jobs in each plan

FLOW PLAN OPEN Open a plan's worktree in a dedicated tmux session (use: flow open)

Switches to or creates a tmux session for the plan’s worktree and opens the interactive status TUI. This provides a one-command entry point into a plan’s interactive environment. If no directory is specified, uses the active job if set.

USAGE flow plan open [directory] [flags]

FLAGS -h, —help help for open

FLOW PLAN RECIPES Manage plan recipes

USAGE flow plan recipes [command]

COMMANDS list List available plan recipes

Flags: -h/—help

Use “flow plan recipes [command] —help” for more information.

FLOW PLAN RECIPES LIST List available plan recipes

USAGE flow plan recipes list [flags]

FLAGS —domain Filter recipes by domain (e.g., generic, grove) -h, —help help for list

FLOW PLAN RESUME Resume a completed interactive agent job

Resumes a completed interactive agent session by finding its native agent session ID and re-launching the agent in a new tmux window.

USAGE flow plan resume <job-file> [flags]

FLAGS -h, —help help for resume

FLOW PLAN REVIEW Mark a plan as ready for review and execute completion hooks (use: flow review)

Marks a plan as ready for review, executes on-review hooks, and prepares it for final cleanup. This is the intermediary step before using ‘flow plan finish’.

USAGE flow plan review [directory] [flags]

FLAGS -h, —help help for review

FLOW PLAN RUN Run jobs (use: flow run)

Run jobs in an orchestration plan. Without arguments, runs the next available jobs. With a single job file argument, runs that specific job. With multiple job file arguments, runs those jobs in parallel.

USAGE flow plan run [job-file…] [flags]

FLAGS -a, —all Run all pending jobs -d, —dir Plan directory (default: .) -h, —help help for run —model Override model for jobs (e.g., claude-3-5-sonnet-20240620, gpt-4) -n, —next Run next available jobs -p, —parallel Max parallel jobs (default: 3) —skip-interactive Skip interactive agent jobs (useful for CI/automation) -w, —watch Watch progress in real-time -y, —yes Skip confirmation prompts

FLOW PLAN SET Set the active job plan directory

Set the active job plan directory to avoid specifying it in every command.

USAGE flow plan set <plan-directory> [flags]

FLAGS -h, —help help for set

EXAMPLES flow plan set user-profile-api flow plan set ./plans/feature-x

FLOW PLAN STATUS Show plan status in an interactive TUI (use: flow status)

Show the status of all jobs in an orchestration plan in an interactive TUI. If no directory is specified, uses the active job if set.

USAGE flow plan status [directory] [flags]

FLAGS -h, —help help for status -t, —tui Launch interactive TUI (default behavior, kept for backwards compatibility)

FLOW PLAN STEP Step through plan execution interactively (use: flow step)

Provides an interactive wizard for executing a plan step by step. Shows runnable jobs and allows you to run, launch, skip, or quit. If no directory is specified, uses the current directory.

USAGE flow plan step [directory] [flags]

FLAGS -h, —help help for step

FLOW PLAN TEMPLATES Manage job templates

USAGE flow plan templates [command]

COMMANDS list List available job templates print Print template contents for agent injection

Flags: -h/—help

Use “flow plan templates [command] —help” for more information.

FLOW PLAN TEMPLATES LIST List available job templates

USAGE flow plan templates list [flags]

FLAGS —domain Filter templates by domain (e.g., generic, grove) -h, —help help for list

FLOW PLAN TEMPLATES PRINT Print template contents for agent injection

Print the contents of a job template.

This is useful for injecting template instructions into a running agent session. By default, only the template body is printed. Use —frontmatter to include the YAML frontmatter.

USAGE flow plan templates print <template-name> [flags]

FLAGS —frontmatter Include YAML frontmatter in output -h, —help help for print

FLOW PLAN TUI Launch interactive TUI for browsing and managing plans

Launch an interactive TUI that provides a navigable view of all plans in your plans directory, similar to ‘flow plan list’ but with interactive features.

Features:

• Navigate through all plans with keyboard (↑/↓, j/k)
• View plan status details (Enter key)
• Execute plan finish command (Ctrl+X)
• Real-time plan list display

USAGE flow plan tui [flags]

FLAGS -h, —help help for tui

FLOW PLAN UNHOLD Remove a plan's on-hold status

Resumes an on-hold plan by removing its status from its .grove-plan.yml file. This makes the plan visible in default views again.

USAGE flow plan unhold [directory] [flags]

FLAGS -h, —help help for unhold

FLOW PLAN UNSET Clear the active job plan directory

Clear the active job plan directory.

USAGE flow plan unset [flags]

FLAGS -h, —help help for unset

FLOW RESUME Resume a completed interactive agent job

Resumes a completed interactive agent session by finding its native agent session ID and re-launching the agent in a new tmux window.

USAGE flow resume <job-file> [flags]

FLAGS -h, —help help for resume

FLOW REVIEW Mark a plan as ready for review and execute completion hooks

Marks a plan as ready for review, executes on-review hooks, and prepares it for final cleanup. This is the intermediary step before using ‘flow finish’.

USAGE flow review [directory] [flags]

FLAGS -h, —help help for review

FLOW RUN Run jobs in an orchestration plan

Run jobs in an orchestration plan. Without arguments, runs the next available jobs. With a single job file argument, runs that specific job. With multiple job file arguments, runs those jobs in parallel.

USAGE flow run [job-file…] [flags]

FLAGS -a, —all Run all pending jobs -d, —dir Plan directory (default: .) -h, —help help for run —model Override model for jobs (e.g., claude-3-5-sonnet-20240620, gpt-4) -n, —next Run next available jobs -p, —parallel Max parallel jobs (default: 3) —skip-interactive Skip interactive agent jobs (useful for CI/automation) -w, —watch Watch progress in real-time -y, —yes Skip confirmation prompts

FLOW SET Set the active job plan directory

Set the active job plan directory to avoid specifying it in every command.

USAGE flow set <plan-directory> [flags]

FLAGS -h, —help help for set

EXAMPLES flow set user-profile-api flow set ./plans/feature-x

FLOW STARSHIP Manage Starship prompt integration

Provides commands to integrate Grove status with the Starship prompt.

USAGE flow starship [command]

COMMANDS install Install the Grove module to your starship.toml

Flags: -h/—help

Use “flow starship [command] —help” for more information.

FLOW STARSHIP INSTALL Install the Grove module to your starship.toml

Appends a custom module to your starship.toml configuration file to display Grove status in your shell prompt. It will also attempt to add the module to your main prompt format.

USAGE flow starship install [flags]

FLAGS -h, —help help for install

FLOW STATUS Show plan status in an interactive TUI

Show the status of all jobs in an orchestration plan within an interactive TUI. If no directory is specified, uses the active job if set. If no active job is set, it will launch the plan browser.

USAGE flow status [directory] [flags]

FLAGS -h, —help help for status -t, —tui Launch interactive TUI (default behavior, kept for backwards compatibility)

FLOW STEP Step through plan execution interactively

Provides an interactive wizard for executing a plan step by step. Shows runnable jobs and allows you to run, launch, skip, or quit. If no directory is specified, uses the current directory.

USAGE flow step [directory] [flags]

FLAGS -h, —help help for step

FLOW TMUX Tmux window management commands

Commands for managing flow in dedicated tmux windows.

USAGE flow tmux [command]

COMMANDS status Open plan status TUI in a dedicated tmux window

Flags: -h/—help

Use “flow tmux [command] —help” for more information.

FLOW TMUX STATUS Open plan status TUI in a dedicated tmux window

Opens the flow plan status TUI in a dedicated tmux window. If the window already exists, it focuses it without disrupting the session. If not in a tmux session, falls back to running the TUI directly.

USAGE flow tmux status [directory] [flags]

FLAGS -h, —help help for status —window-name Name of the tmux window (default: plan)

FLOW UNHOLD Remove a plan's on-hold status

Resumes an on-hold plan by removing its status from its .grove-plan.yml file. This makes the plan visible in default views again.

USAGE flow unhold [directory] [flags]

FLAGS -h, —help help for unhold

FLOW UNSET Clear the active job plan directory

Clear the active job plan directory.

USAGE flow unset [flags]

FLAGS -h, —help help for unset

FLOW VERSION Print the version information for this binary

USAGE flow version [flags]

FLAGS -h, —help help for version —json Output version information in JSON format