CLI Reference

Complete reference for every kendr subcommand and flag. Every command produces colored, structured terminal output via rich. Use --json for machine-readable output and --quiet to suppress progress messages.

Global Flags

These flags apply to the kendr root command and must appear before the subcommand.

Flag	Description
`--no-color`	Disable ANSI colors in all CLI output.
`--log-level LEVEL`	Global log verbosity hint for runtime services. Choices: `silent`, `fatal`, `error`, `warn`, `info`, `debug`, `trace`. Sets `KENDR_LOG_LEVEL` environment variable for the run.
`-V` / `--version`	Print the current version and exit.

`kendr run`

Run the multi-agent orchestrator for a single query. This is the primary entry point for all intelligence workflows.

The gateway is NOT auto-started. If your query requires the gateway, start it first with kendr gateway start.

bash


kendr run [FLAGS] "your query"

Core Flags

Flag	Default	Description
`query`	_(positional)_	The query or task to execute.
`--max-steps N`	`20`	Maximum orchestration steps before the run ends.
`--working-directory PATH`	_(from env)_	Directory for output artifacts.
`--current-folder`	_(off)_	Use the current terminal working directory as output.
`--auto-approve`	_(off)_	Auto-approve all blueprint and plan gates without interactive prompts.
`--skip-reviews`	_(off)_	Skip reviewer-agent checks between orchestration steps.
`--max-step-revisions N`	`0`	Override maximum reviewer revisions per step (0 = agent default).
`--json`	_(off)_	Emit the final run state as JSON instead of rich output.
`--quiet`	_(off)_	Suppress live step progress messages.

Research and Sources

Flag	Default	Description
`--sources LIST`	_(auto)_	Comma-separated list of research sources for the multi-source pipeline. Options: `web`, `arxiv` (alias `papers`, `academic`), `reddit` (alias `social`), `scholar`, `patents` (alias `patent`), `openalex`, `local` (requires `--drive`). Example: `--sources arxiv,reddit,web`
`--pages N`	`0`	Target page count for long-form document output. Implies `--long-document` mode.
`--research-model MODEL`	_(env)_	Override the deep-research model for this run. Example: `o4-mini-deep-research`.
`--research-instructions TEXT`	_(none)_	Extra instructions appended to every deep-research call.
`--research-max-wait-seconds N`	`0`	Max seconds to wait per deep-research API call before timeout handling.
`--research-poll-interval-seconds N`	`0`	Polling interval for background deep-research status checks.
`--research-max-tool-calls N`	`0`	Maximum web tool calls per deep-research pass.
`--research-max-output-tokens N`	`0`	Optional output token cap per deep-research pass.

Long-Form Document Mode

Flag	Default	Description
`--long-document`	_(off)_	Force staged long-form document workflow (chaptered research + merged output).
`--long-document-pages N`	`0`	Target page count for long-form output.
`--long-document-sections N`	`0`	Explicit chapter/section count.
`--long-document-section-pages N`	`0`	Approximate pages per section.
`--long-document-title TITLE`	_(auto)_	Override the report title.
`--long-document-no-collect-sources`	_(off)_	Skip the pre-collection evidence bank step.
`--long-document-no-section-search`	_(off)_	Skip per-section web search.
`--long-document-section-search-results N`	`0`	Web search results to gather per section.
`--long-document-no-visuals`	_(off)_	Skip generating tables and flowcharts.

Local Drive Intelligence

Flag	Default	Description
`--drive PATH`	_(none)_	Local folder or file to ingest. Repeat for multiple paths.
`--drive-min-files N`	`0`	Minimum file count for long-form report mode.
`--drive-max-files N`	`0`	Maximum files to process.
`--drive-extensions LIST`	_(all)_	Comma-separated extension allowlist (example: `pdf,docx,xlsx`).
`--drive-no-recursive`	_(off)_	Disable recursive folder traversal.
`--drive-include-hidden`	_(off)_	Include hidden files and folders.
`--drive-disable-image-ocr`	_(off)_	Disable OCR for images in drive paths.
`--drive-ocr-instruction TEXT`	_(auto)_	Custom OCR instruction override.
`--drive-no-memory-index`	_(off)_	Skip vector-memory indexing of extracted summaries.
`--drive-auto-generate-extension-handlers`	_(off)_	Enable dynamic agent generation for unsupported file types.

Dev Pipeline Mode (`--dev`)

Activates the end-to-end dev pipeline: blueprint → scaffold → build → test → verify (with auto-fix) → zip export. Equivalent to kendr generate with full pipeline orchestration.

Flag	Default	Description
`--dev`	_(off)_	Activate the dev pipeline mode.
`--stack TEMPLATE`	_(LLM selects)_	Tech stack template. Options: `fastapi_postgres`, `fastapi_react_postgres`, `nextjs_prisma_postgres`, `express_prisma_postgres`, `mern_microservices_mongodb`, `pern_postgres`, `nextjs_static_site`, `django_react_postgres`, `custom_freeform`.
`--dev-skip-tests`	_(off)_	Skip test generation and execution.
`--dev-skip-devops`	_(off)_	Skip Dockerfile, docker-compose, and CI/CD generation.
`--dev-max-fix-rounds N`	`3`	Maximum auto-fix rounds when the verifier fails.

Codebase Analysis

Flag	Default	Description
`--codebase`	_(off)_	Analyze an existing codebase before planning.
`--codebase-path PATH`	_(working dir)_	Path to the existing project.
`--codebase-max-files N`	`0`	Maximum files to scan (0 = 1000).

superRAG Knowledge Engine

Flag	Default	Description
`--superrag-mode MODE`	_(none)_	superRAG operating mode: `build`, `chat`, `switch`, `list`, `status`.
`--superrag-session ID`	_(none)_	Session ID to reuse or switch to.
`--superrag-new-session`	_(off)_	Force creation of a new session for build mode.
`--superrag-session-title TITLE`	_(auto)_	Human-readable title for the session.
`--superrag-path PATH`	_(none)_	Local path to ingest. Repeat for multiple.
`--superrag-url URL`	_(none)_	Seed URL to crawl and ingest. Repeat for multiple.
`--superrag-db-url URL`	_(none)_	Database URL for schema and row-sample ingestion.
`--superrag-db-schema SCHEMA`	_(auto)_	Optional database schema to target.
`--superrag-onedrive`	_(off)_	Enable OneDrive ingestion (requires Microsoft Graph).
`--superrag-onedrive-path PATH`	_(root)_	OneDrive folder path to ingest.
`--superrag-chat QUESTION`	_(none)_	Question to ask in chat mode.
`--superrag-top-k N`	`0`	Top-K vector matches for chat (0 = agent default).

Coding Workflows

Flag	Default	Description
`--coding-context-file FILE`	_(none)_	Project file to load as coding context. Repeat for multiple.
`--coding-write-path PATH`	_(none)_	Target file for coding output.
`--coding-instructions TEXT`	_(none)_	Extra instructions for coding agents.
`--coding-language LANG`	_(auto)_	Coding language hint.
`--coding-backend BACKEND`	`auto`	Preferred code-generation backend: `auto`, `codex-cli`, `openai-sdk`, `responses-http`.

Local Command Execution

Flag	Default	Description
`--os-command CMD`	_(none)_	Execute one explicit command through `os_agent`.
`--os-shell SHELL`	_(auto)_	Preferred shell: `bash`, `powershell`, `cmd`.
`--os-timeout N`	`0`	Command timeout in seconds.
`--os-working-directory PATH`	_(working dir)_	Working directory for command execution.
`--target-os OS`	_(auto)_	Target OS hint: `linux`, `macos`, `windows`.

Privileged Mode

Flag	Default	Description
`--privileged-mode`	_(off)_	Enable privileged policy controls for this run.
`--privileged-approved`	_(off)_	Confirm explicit operator approval for privileged actions. Required together with `--privileged-approval-note`.
`--privileged-approval-note NOTE`	_(none)_	Ticket or approval reference for auditing.
`--privileged-read-only`	_(off)_	Force read-only execution; no writes or mutations.
`--privileged-allow-root`	_(off)_	Allow sudo/root escalation.
`--privileged-allow-destructive`	_(off)_	Allow destructive operations.
`--privileged-enable-backup`	_(off)_	Create filesystem snapshot before mutating actions.
`--privileged-allowed-path PATH`	_(none)_	Allowed path root. Repeat for multiple.
`--privileged-allowed-domain DOMAIN`	_(none)_	Allowed network domain. Repeat for multiple.
`--kill-switch-file FILE`	_(env)_	Halt if this file exists before next agent step.

Security Workflows

Flag	Default	Description
`--security-authorized`	_(off)_	Confirm you are explicitly authorized to run defensive security tasks on the target.
`--security-target-url URL`	_(none)_	Target URL for security assessment.
`--security-authorization-note NOTE`	_(none)_	Ticket/approval reference proving assessment authorization.
`--security-scan-profile PROFILE`	`standard`	Scan depth: `baseline`, `standard`, `deep`, `extensive`.
`--no-auto-install-security-tools`	_(off)_	Disable automatic installation of missing security tools (nmap, zap).

Communication

Flag	Default	Description
`--communication-authorized`	_(off)_	Confirm authorization to access communication channels for this run.
`--communication-lookback-hours N`	`24`	Lookback window in hours for the communication digest.
`--whatsapp-to PHONE`	_(none)_	Recipient phone in E.164 format for `whatsapp_send_message_agent`.
`--whatsapp-message TEXT`	_(none)_	Plain text message body.
`--whatsapp-template NAME`	_(none)_	WhatsApp template name.
`--whatsapp-template-language CODE`	`en_US`	Language code for the template.

Session Continuity

Flag	Default	Description
`--channel ID`	_(none)_	Channel ID for conversational session continuity (e.g. `webchat`, `slack`).
`--workspace-id ID`	_(none)_	Workspace ID for session routing.
`--sender-id ID`	_(none)_	Sender/user ID for session routing.
`--chat-id ID`	_(none)_	Chat/thread ID for session routing.
`--session-key KEY`	_(none)_	Explicit session key in format `channel:workspace:chat:scope`.
`--new-session`	_(off)_	Force creation of a fresh session for this run.

Examples

bash


# Research brief
kendr run --current-folder "Analyze Stripe: business model, competitors, and key risks."

# Multi-source research into a 20-page document
kendr run --sources arxiv,web,reddit --pages 20 "AI safety landscape 2024"

# superRAG: build a knowledge base
kendr run --superrag-mode build --superrag-new-session --superrag-session-title "docs_kb" \
  --superrag-path ./docs --superrag-url https://example.com/help "Index our knowledge base."

# superRAG: chat with a session
kendr run --superrag-mode chat --superrag-session docs_kb \
  --superrag-chat "What are the installation requirements?"

# Dev pipeline with stack template
kendr run --dev --stack fastapi_postgres "Build a task management API"

# Privileged OS command
kendr run --current-folder --os-command "ls -la" --os-shell bash \
  --privileged-approved --privileged-approval-note "OPS-42 approved" "List project root."

# Communication digest (last 8 hours)
kendr run --communication-authorized \
  "Summarize my Slack and Gmail messages from the last 8 hours."

`kendr research`

Run the multi-source research pipeline and generate a document directly. A focused shortcut for research-specific workflows.

bash


kendr research [FLAGS] "research topic"

Flag	Default	Description
`query`	_(positional)_	Research query or topic.
`--sources LIST`	_(auto)_	Comma-separated research sources: `web`, `arxiv`, `reddit`, `scholar`, `patents`, `openalex`, `local`.
`--pages N`	`0`	Target page count. Enables long-document mode automatically.
`--title TEXT`	_(auto)_	Optional report title.
`--drive PATH`	_(none)_	Local path to include as a research source. Repeat for multiple.
`--research-model MODEL`	_(env)_	Override the deep-research model.
`--auto-approve`	_(off)_	Auto-approve plan gates.
`--max-steps N`	`20`	Maximum orchestration steps.
`--working-directory PATH`	_(env)_	Output directory.
`--current-folder`	_(off)_	Use current terminal folder as output.
`--json`	_(off)_	Emit final state as JSON.
`--quiet`	_(off)_	Suppress progress messages.

Examples

bash


# Quick research brief
kendr research "Battery recycling market: key players and investment risks"

# Multi-source with document output
kendr research --sources arxiv,web --pages 15 --title "Quantum ML Survey 2024" \
  "Survey of quantum machine learning advances"

# Include local files as a source
kendr research --sources local,web --drive ./internal-reports \
  "Summarize our internal research and compare with market trends"

`kendr generate`

Generate a complete multi-agent software project from a description: blueprint → scaffold → build → test → verify → zip export.

bash


kendr generate [FLAGS] "project description"

Flag	Default	Description
`description`	_(positional)_	Natural language project description.
`--name NAME`	_(auto)_	Project name in kebab-case. Derived from description if omitted.
`--stack TEMPLATE`	_(LLM selects)_	Tech stack template: `fastapi_postgres`, `fastapi_react_postgres`, `nextjs_prisma_postgres`, `express_prisma_postgres`, `mern_microservices_mongodb`, `pern_postgres`, `nextjs_static_site`, `django_react_postgres`, `custom_freeform`.
`--output PATH`	_(working dir)_	Output directory root.
`--auto-approve`	_(off)_	Auto-approve blueprint and plan gates.
`--skip-tests`	_(off)_	Skip test generation and execution.
`--skip-devops`	_(off)_	Skip Dockerfile, docker-compose, and CI/CD generation.
`--skip-reviews`	_(off)_	Skip reviewer agent checks between build steps.
`--max-steps N`	`40`	Maximum orchestration steps.
`--working-directory PATH`	_(env)_	Working directory for artifacts.
`--current-folder`	_(off)_	Use current terminal folder.
`--json`	_(off)_	Emit final state as JSON.
`--quiet`	_(off)_	Suppress progress messages.

Examples

bash


# Generate a FastAPI + PostgreSQL API
kendr generate --stack fastapi_postgres "A task management REST API with auth and tests"

# Full-stack with auto-approve
kendr generate --stack fastapi_react_postgres --auto-approve \
  "SaaS starter with user auth, billing, and an admin dashboard"

# Simple script project (LLM chooses stack)
kendr generate "A Python CLI tool for batch image resizing"

`kendr gateway`

Start, stop, restart, or inspect the HTTP gateway server. The gateway must be started explicitly before any workflow that requires it. It is never auto-started by kendr run, kendr research, or kendr generate.

bash


kendr gateway start
kendr gateway stop
kendr gateway restart
kendr gateway status [--json]
kendr gateway serve       # run in foreground (default when no action given)

Action	Description
`start`	Start the gateway server in the background. Writes PID to `~/.kendr/gateway.pid`.
`stop`	Stop the gateway-owned process on the configured port. Does not kill unrelated processes.
`restart`	Stop the current gateway and start a fresh one.
`status`	Show gateway health, PID state, and port listener status.
`serve`	Run the gateway in the foreground (default when no action is specified).

Flag	Default	Description
`--json`	_(off)_	Emit machine-readable status as JSON (includes `health_ok`, `pid_alive`, `pid_owned`, `port_listening`).

Examples

bash


kendr gateway start
kendr gateway status
kendr gateway stop
kendr gateway restart
kendr gateway status --json
kendr gateway serve        # foreground — useful for debugging

`kendr web`

Alias for kendr gateway serve. Runs the gateway server in the foreground.

bash


kendr web

`kendr setup-ui`

Run the OAuth and setup UI directly in the foreground (shortcut for kendr setup ui).

bash


kendr setup-ui

`kendr setup`

Manage component configuration via the CLI. Use kendr setup status after any configuration change.

bash


kendr setup <action> [args]

Action	Description
`status`	Show setup status: which components are configured, available agents, and setup actions.
`components`	List all configurable components (integrations).
`show COMPONENT`	Show one component's current configuration.
`set COMPONENT KEY VALUE`	Set one configuration field for a component (stored in local DB). Pass `--secret` to mark the value as a secret (accepted for compatibility — secret classification is otherwise catalog-driven).
`unset COMPONENT KEY`	Remove one configuration field.
`enable COMPONENT`	Enable a component.
`disable COMPONENT`	Disable a component.
`export-env`	Export current DB configuration as dotenv lines. Use `--include-secrets` to include secrets.
`install`	Install auto-installable local components (nmap, zap, dependency-check, playwright). Use `--yes` to skip confirmation, `--only` to install specific tools.
`ui`	Run the web-based OAuth setup UI on `http://127.0.0.1:8787`.
`oauth PROVIDER`	Run the OAuth login flow for a supported provider (`google`, `microsoft`, `slack`, `all`). Use `--no-browser` to print URLs without opening a browser. Use `--ensure-ui` for compatibility with environments that auto-start the setup UI.

Examples

bash


# Minimum viable setup
kendr setup set core_runtime KENDR_WORKING_DIR /home/user/kendr-work
kendr setup set openai OPENAI_API_KEY sk-...
kendr setup status

# Check one component
kendr setup show openai --json

# Install security tools
kendr setup install --yes

# Install only specific tools
kendr setup install --yes --only nmap zap

# Run OAuth flow without opening a browser (print URL to copy manually)
kendr setup oauth google --no-browser

# Export config for a .env file
kendr setup export-env > .env

`kendr resume`

Resume or inspect a previously run or paused orchestration session.

bash


kendr resume [FLAGS] [run-folder]

Flag	Default	Description
`target`	_(positional)_	Run folder, manifest/checkpoint file, or working directory path.
`query`	_(positional)_	Optional reply or new query when resuming or branching.
`--output-folder PATH`	_(none)_	Explicit path to the run output folder or manifest.
`--working-directory PATH`	_(none)_	Working directory to search for persisted run folders.
`--latest`	_(off)_	Automatically use the newest discovered run candidate.
`--inspect`	_(off)_	Inspect the session without executing it.
`--branch`	_(off)_	Start a new child run from the saved context instead of resuming in place.
`--reply TEXT`	_(none)_	Explicit reply for a run paused at a user-input gate.
`--force`	_(off)_	Take over a run marked as running or stale.
`--json`	_(off)_	Emit resume candidate or result as JSON.

Examples

bash


# Inspect the most recent run in the current folder
kendr resume --working-directory . --latest --inspect

# Resume a specific run
kendr resume --output-folder ./output/runs/run_cli_abc123

# Reply to a paused approval gate
kendr resume --output-folder ./output/runs/run_cli_abc123 --reply approve

# Branch: start a new run from a completed session
kendr resume --output-folder ./output/runs/run_cli_abc123 --branch \
  "Expand the brief into an investor-facing memo."

`kendr agents`

List or inspect discovered agents.

bash


kendr agents list [FLAGS]
kendr agents show AGENT_NAME [--json]

Flag	Description
`--plugin NAME`	Filter by plugin name.
`--contains TEXT`	Filter agent name/description by substring.
`--limit N`	Limit number of results.
`--json`	Emit as JSON.

Examples

bash


kendr agents list
kendr agents show deep_research_agent --json
kendr agents list --contains research

`kendr plugins`

List discovered plugins (built-in and external).

bash


kendr plugins list [FLAGS]

Flag	Description
`--kind KIND`	Filter by plugin kind.
`--contains TEXT`	Filter by name/description substring.
`--limit N`	Limit results.
`--json`	Emit as JSON.

`kendr sessions`

List and manage conversational sessions.

bash


kendr sessions list [FLAGS]
kendr sessions current
kendr sessions use SESSION_KEY
kendr sessions clear

Flag	Description
`--limit N`	Number of sessions to list (default 20).
`--json`	Emit as JSON.

`kendr workdir`

Manage the configured working directory.

bash


kendr workdir show [--json]
kendr workdir set PATH
kendr workdir here
kendr workdir create PATH [--activate]

Subcommand	Description
`show`	Show the currently configured working directory.
`set PATH`	Set the working directory to an existing absolute path.
`here`	Set the current terminal folder as the working directory.
`create PATH`	Create a new working directory. Use `--activate` to set it active immediately.
`clear`	Clear the configured working directory (unset `KENDR_WORKING_DIR`).

Examples

bash


kendr workdir here
kendr workdir show
kendr workdir set /home/user/my-project

`kendr status`

Show a runtime status snapshot (agents available, setup health, session state).

bash


kendr status [--json]

`kendr daemon`

Run the always-on monitor and heartbeat loop for scheduled tasks and stock/news monitoring.

bash


kendr daemon [--poll-interval N] [--heartbeat-interval N] [--once]

Flag	Default	Description
`--poll-interval N`	_(env)_	Seconds between monitor passes.
`--heartbeat-interval N`	_(env)_	Seconds between heartbeat log entries.
`--once`	_(off)_	Run one monitor pass and exit.

`kendr rollback`

List or apply filesystem snapshots created by privileged runs.

bash


kendr rollback list
kendr rollback apply --snapshot PATH [--target-dir PATH] [--overwrite] [--yes]

`kendr hello`

Display a quick-start welcome screen with setup guidance and example commands. Useful for first-time orientation.

bash


kendr hello [--json]

Flag	Default	Description
`--json`	_(off)_	Emit quick-start info as JSON.

`kendr help`

Show help for a specific command.

bash


kendr help [TOPIC]

Argument	Description
`topic`	Optional command name to show help for. Omit to show general help.

Examples

bash


kendr help
kendr help run
kendr help gateway

`kendr mcp`

Manage MCP (Model Context Protocol) server connections. Kendr acts as an MCP client — any tools exposed by registered servers are automatically discovered and made available to the agent runtime.

bash


kendr mcp <action> [args]

Subcommands

Subcommand	Description
`list`	List all registered MCP servers and their discovered tools.
`add`	Register a new HTTP or stdio MCP server.
`remove`	Remove a registered server by name or ID.
`discover`	Re-run tool discovery on a registered server.
`test`	Ping a server and list its available tools.
`enable`	Enable a disabled server so kendr uses its tools.
`disable`	Disable a server without removing it.
`zapier`	Quick-connect your Zapier MCP server (see below).

`kendr mcp list`

List all registered servers with their connection type, status, tool count, and ID.

bash


kendr mcp list

`kendr mcp add`

bash


kendr mcp add <name> <connection>
kendr mcp add --name <name> --url <url> [--type http|stdio] [--auth-token TOKEN] [--description TEXT] [--no-discover]

Flag	Default	Description
`name`	_(positional)_	Human-readable server name.
`connection`	_(positional)_	HTTP endpoint URL or shell command (for stdio servers).
`--name NAME`	—	Server name (overrides positional).
`--url URL`	—	Connection (overrides positional).
`--type`	`http`	Connection type: `http` (HTTP/SSE endpoint) or `stdio` (shell command).
`--auth-token TOKEN`	—	Bearer token sent as `Authorization: Bearer …` for authenticated HTTP servers.
`--description TEXT`	—	Optional description stored with the server entry.
`--no-discover`	_(off)_	Register without immediately running tool discovery.

#### Examples

bash


# Add an HTTP MCP server
kendr mcp add "My Server" http://localhost:8000/mcp

# Add with auth token
kendr mcp add "Private Server" https://api.example.com/mcp --auth-token sk-abc123

# Add a stdio server
kendr mcp add "Local Agent" "python mcp_servers/my_server.py" --type stdio

# Register without discovering tools yet
kendr mcp add "Slow Server" http://slow-server.internal/mcp --no-discover

`kendr mcp remove`

Remove a registered MCP server and all its discovered tool metadata.

bash


kendr mcp remove <name-or-id>

`kendr mcp discover`

Re-connect to a registered server and refresh its tool list. Useful after the server is updated.

bash


kendr mcp discover <name-or-id>

`kendr mcp test`

Ping a server and print all its available tools without modifying the registry.

bash


kendr mcp test <name-or-id>

`kendr mcp enable` / `kendr mcp disable`

Enable or disable a server. Disabled servers are skipped during agent runtime but remain in the registry.

bash


kendr mcp enable <name-or-id>
kendr mcp disable <name-or-id>

`kendr mcp zapier`

Quick-connect a Zapier MCP server. Zapier exposes 7,000+ app integrations (Gmail, Slack, Notion, GitHub, Salesforce, and more) through a single personal MCP endpoint.

bash


kendr mcp zapier <your-zapier-mcp-url>
kendr mcp zapier --url <url> [--name NAME] [--no-discover]

Flag	Default	Description
`mcp_url`	_(positional)_	Your personal Zapier MCP URL.
`--url URL`	—	Alternative to positional argument.
`--name NAME`	`Zapier`	Name for this registry entry.
`--no-discover`	_(off)_	Register without running discovery immediately.

Get your Zapier MCP URL:

1. Go to zapier.com/mcp 2. Copy your personal MCP endpoint URL 3. It will look like: https://mcp.zapier.com/api/mcp/s/<token>/mcp

#### Examples

bash


# Connect Zapier MCP (tool discovery runs automatically)
kendr mcp zapier https://mcp.zapier.com/api/mcp/s/abc123xyz/mcp

# Connect with a custom name
kendr mcp zapier https://mcp.zapier.com/api/mcp/s/abc123xyz/mcp --name "My Zapier"

# Connect without immediate discovery
kendr mcp zapier https://mcp.zapier.com/api/mcp/s/abc123xyz/mcp --no-discover

Running this command without a URL prints setup instructions:

bash


kendr mcp zapier

Shell Automation

Kendr's shell automation lets agents install tools, run multi-step OS commands, and execute complex workflows directly on your machine — all from the chat UI or CLI.

How it works

When shell automation is active, the OS agent (os_agent) and the shell plan agent (shell_plan_agent) are authorised to:

- Run arbitrary shell commands (bash, sh, POSIX-compatible) - Install packages with apt, brew, pip, npm, cargo, go install, pipx, etc. - Start and manage services with docker, docker compose, systemctl, ollama, nginx, etc. - Clone repos, pull updates, and run build pipelines with git, make, cmake, etc.

For multi-step goals (e.g. "install Docker, pull an image, and run a container"), Kendr uses shell_plan_agent which:

1. Decomposes the goal into ordered steps via LLM 2. For each step, runs an optional check command first — if the check passes, the step is skipped (idempotent) 3. Executes the command and captures stdout/stderr 4. Reports per-step status: ✓ done / skipped / blocked / error

Enabling Shell Automation in the Chat UI

Click the 💻 Shell button in the chat header. When it turns amber the mode is active. A warning banner appears below the input box as a reminder that agents may run commands on your machine.

To disable, click the button again.

Enabling Shell Automation via CLI

Pass --privileged-approved together with a note to any kendr run invocation:

bash


kendr run "Install nginx, configure it, and start it" \
  --privileged-approved \
  --privileged-approval-note "Approved by me for nginx setup"

Or use the environment variable to auto-approve for all shell commands in a session:

bash


export KENDR_AUTO_APPROVE_COMMANDS=true
kendr run "Set up a Python venv and install fastapi uvicorn"

Policy and safety

Shell automation follows kendr's privileged control policy:

Situation	Behaviour
Non-destructive command, auto-approve ON	Runs automatically
Destructive command (rm -rf, DROP TABLE, etc.)	Always prompts — never auto-approved
`sudo` or root-requiring command	Blocked unless `--privileged-allow-root` is set
Path outside working directory	Blocked unless `--privileged-allowed-paths` includes it

To allow destructive commands explicitly:

bash


kendr run "Clean build artefacts" \
  --privileged-approved \
  --privileged-approval-note "Safe to delete build/" \
  --privileged-allow-destructive

Terminal output rendering

Shell automation results in the Chat UI are rendered in a dark terminal style with colour-coded lines:

- Green — command being run - Blue-grey — standard output - Red — errors / stderr - Amber — blocked commands - Muted — skipped steps

Example goals

bash


# Install and start Ollama with llama3
kendr run "Install ollama, pull llama3, and start the server" --privileged-approved --privileged-approval-note "Local dev setup"

# Dockerise a project
kendr run "Write a Dockerfile for this Python project and build it" --privileged-approved --privileged-approval-note "Build image"

# Set up nginx reverse proxy
kendr run "Install nginx and configure it to proxy port 3000 on /app" --privileged-approved --privileged-approval-note "Dev proxy"

CLI Reference

Global Flags

kendr run

Core Flags

Research and Sources

Long-Form Document Mode

Local Drive Intelligence

Dev Pipeline Mode (--dev)

Codebase Analysis

superRAG Knowledge Engine

Coding Workflows

Local Command Execution

Privileged Mode

Security Workflows

Communication

Session Continuity

Examples

kendr research

Examples

kendr generate

Examples

kendr gateway

Examples

kendr web

kendr setup-ui

kendr setup

Examples

kendr resume

Examples

kendr agents

Examples

kendr plugins

kendr sessions

kendr workdir

Examples

kendr status

kendr daemon

kendr rollback

kendr hello

kendr help

Examples

kendr mcp

Subcommands

kendr mcp list

kendr mcp add

kendr mcp remove

kendr mcp discover

kendr mcp test

kendr mcp enable / kendr mcp disable

kendr mcp zapier

Shell Automation

How it works

Enabling Shell Automation in the Chat UI

Enabling Shell Automation via CLI

Policy and safety

Terminal output rendering

Example goals

`kendr run`

Dev Pipeline Mode (`--dev`)

`kendr research`

`kendr generate`

`kendr gateway`

`kendr web`

`kendr setup-ui`

`kendr setup`

`kendr resume`

`kendr agents`

`kendr plugins`

`kendr sessions`

`kendr workdir`

`kendr status`

`kendr daemon`

`kendr rollback`

`kendr hello`

`kendr help`

`kendr mcp`

`kendr mcp list`

`kendr mcp add`

`kendr mcp remove`

`kendr mcp discover`

`kendr mcp test`

`kendr mcp enable` / `kendr mcp disable`

`kendr mcp zapier`