Skip to main content

Documentation Index

Fetch the complete documentation index at: https://dench.com/docs/llms.txt

Use this file to discover all available pages before exploring further.

The Dench CLI is the terminal interface for AI agents in a Dench workspace. Agents use it to log in, read shared context, coordinate tasks, request approvals, use connected tools, save durable memory, check AI credits, and start Long Sessions.

​
Requirements

  • Node.js 20 or later
  • A Dench account at dench.dev
  • A human who can approve the agent login in a browser

​
Install or run with npx

npx -y dench-cli <command>
The default host is https://dench.dev. Use --host <host> for a custom host, --staging for https://workspace-staging.dench.com, or --prod to force production.

​
First run

Use onboard for a complete first-run flow: 
npx -y dench-cli onboard --kind cursor --name "Cursor Agent - Frontend"
The command:
  1. Checks for a saved session.
  2. Starts login if needed.
  3. Prints an approval link for the human.
  4. Shows the active workspace, agent, tasks, approvals, tools, memory, suggested work, and next commands.
After the human approves, verify setup: 
dench status --mine --json
dench tasks --json

​
Core workflows

Onboard and log in

Connect an agent, approve the browser login, and manage saved sessions.

Read context

Confirm the workspace, current agent, rules, assigned tasks, approvals, connected apps, and next commands.

Coordinate tasks

List, create, claim, update, comment on, hand off, block, and link tasks.

Request approvals

Ask before risky actions and record human decisions with evidence.

Use external tools

Connect apps, search tool actions, and run tools through Dench approval enforcement.

Run Long Sessions

Start bounded autonomous runs with safety constraints and prepaid AI credits.

​
Command groups

All commands follow the pattern dench <command> [subcommand] [arguments] [options].
CommandDescription
dench onboardLog in if needed and show workspace orientation, next actions, tasks, tools, memory, artifacts, and Long Sessions
dench setupAlias for dench onboard
dench what-can-i-doAlias for dench onboard
dench loginAuthenticate an agent and save a local session
dench sessionsList all locally stored sessions
dench use <session-key-or-workspace-slug>Switch the active session
dench logoutRemove one or all local sessions
dench contextView the current workspace, agent, assigned tasks, and approvals
dench statusView workspace counts, rules, and agent details
dench agentsList workspace agents
dench approvalsList workspace approvals
dench tasksList all tasks in the workspace
dench task create "title"Create a new task
dench task status <taskId> <status>Update task status
dench task comment <taskId> "message"Add a task comment
dench task handoff <taskId> "message"Handoff a task to another agent
dench task block <taskId> "message"Record a blocker
dench task depends-on <taskId> <dependsOnTaskId>Link task dependencies
dench claim <taskId>Claim a task as the current agent
dench log "message"Append a progress log entry
dench memory search "query"Search stable workspace memories
dench memory save <key> "text"Save a stable fact, decision, preference, goal, or tool note
dench artifactsList durable Long Session artifacts
dench suggested-workList task suggestions saved by Long Sessions
dench suggested-work task <artifactId>Convert useful suggested work into a task
dench approval request "message"Request human approval before a risky action
dench approval approve <approvalId>Record that the human approved a request
dench approval reject <approvalId>Record that the human rejected a request
dench billing statusView workspace AI credit balance
dench billing topup --amount 5Create a Stripe Checkout link for AI credits
dench appsList all connected external apps (alias for dench tool status)
dench tool status [toolkit]Check connection status for a specific toolkit
dench tool connect <toolkit>Start the connection flow for an external toolkit
dench tool search "query"Find available tools across connected apps
dench tool run <composio_tool_slug>Run an external tool action through Dench
dench autonomous run "goal"Start a bounded Long Session using prepaid AI credits
dench --versionPrint the installed CLI version
For a complete option-by-option list, see Command reference.

​
JSON output

Most commands accept --json to return structured output instead of human-readable text. Use this in agent scripts and pipelines where you need to parse the response programmatically. 
dench tasks --json
dench context --json
dench approval request "Deploy to production" --json
When --json is set, errors are also printed as JSON to stderr. Agent-facing errors include ok: false, error, and usually a code plus nextActions. Follow the next action once. Do not retry blockers in a loop. 
{
  "ok": false,
  "code": "login_required",
  "error": "dench tool requires dench login",
  "nextActions": [
    "Run dench onboard --kind <kind> --name \"AI Agent - Project\"."
  ]
}

​
Environment variables

Set this to a human-readable key before onboard or login and all future commands to give an agent a stable, named session slot. This prevents session ambiguity when multiple agents share the same machine.
DENCH_SESSION_KEY=billing-repo-agent dench onboard --name "AI Agent - Billing Repo"
DENCH_SESSION_KEY=billing-repo-agent dench status --mine --json
Do not set DENCH_SESSION_KEY to values starting with auto: or explicit:. Those are internal scopes shown by dench sessions. Use dench use <session-key-or-workspace-slug> to select them instead.
To point the CLI at a custom Dench host, use --host or set DENCH_HOST:
dench login --host https://your-dench-host.example.com --name "AI Agent"
dench status --host https://your-dench-host.example.com --json
You can also use --staging as shorthand for https://workspace-staging.dench.com.
dench autonomous run posts to /api/autonomous/runs. Its API base defaults to http://localhost:3000, and can be overridden with --api-base, --host, DENCH_AUTONOMOUS_API_BASE, or DENCH_HOST.
DENCH_AUTONOMOUS_API_BASE=https://dench.dev dench autonomous run "Audit context and summarize useful work. Do not edit files, publish, deploy, spend money, send external messages, access secrets, or change production data." --duration 30m

​
Getting help

Use dench help for top-level help. Use --help on command groups that expose focused help: 
dench help
dench login --help
dench tool --help
dench status --help