Skip to content
NewApprove, edit or reject from the keyboard

Hand off the work.
Keep the final say.

Dispatch is a Kanban board where every card is a task for an AI agent. Watch it work step by step, review what it made, and approve anything that leaves the board.

Press Enter to open the demo with this task. Shift and Enter adds a new line.

Try

Opens a private demo workspace · 3 free runs · no sign-up. Or start free

[01]Product

Delegation you can actually watch.

Every card is a task, every task gets an agent, and every step is on the record, from the first search to the last cent.

Open the live demo

Live preview · three agents, one board

  • Watch every step

    A live trace shows the plan, each tool call, each source and the cost, while it happens.

  • Approve what matters

    Emails, Slack posts and webhooks wait for you. Approve, edit or reject with one key.

  • Run cards in parallel

    Drag three cards to Running and three agents start at once. Up to five at a time on Pro.

  • Cited, exportable results

    Documents, tables, JSON and email drafts, with a source behind every claim.

[02]Approvals

Nothing leaves the board without you.

When an agent wants to send an email, post to Slack or call a webhook, the run pauses and the card shows Awaiting approval. You see the exact action before it happens.

  1. A
    ApproveThe action runs exactly as shown, and the run carries on.
  2. E
    Edit, then approveChange the payload first. The diff is kept with the decision.
  3. R
    Reject with a reasonThe agent proposes an alternative, or finishes without the action.

Every decision is logged: who, when and what changed.How approvals work

Approval needed

Draft a follow-up email for Jane · run 1 · step 4

Awaiting approval
Tool
send_emailside effect
To
jane@northwind.example
Subject
Following up on the Northwind pilot

Hi Jane,

Thanks again for Tuesday. As promised, here is a two-week pilot for the ops team: three boards, forty runs and a short review every Friday. Would Thursday at 10:00 work for a 20-minute kickoff?

Best, Nadia

The run is paused until someone decides.

[03]How it works

From a sentence to a signed-off result.

Three steps, the same every time. The agent does the legwork; you keep the decisions.

  1. [01]

    Describe it.

    Write the task in plain words or pick a task type. Attach files, add links, choose a reviewer. It becomes a card.

  2. [02]

    Watch it work.

    Drag the card to Running. The agent plans, then calls tools you allowed, and every step streams into the trace.

  3. [03]

    Approve the result.

    The artifact lands in Review with its sources. Anything that would leave the board waits for your yes.

[04]Task types

Five task types, ready on day one.

Each type knows its inputs, the tools it may use, what it produces and when it must ask you. Pro adds your own.

Try them in the demo

  • Research brief

    Searches the web, reads the best sources and writes a cited brief with a summary and key findings.

    Inputs
    Topic or question · Depth · Language
    Tools
    web_searchfetch_urlread_file
    Output
    Document

  • Competitor summary

    Reads competitor sites and summarizes pricing, features and positioning, with a comparison table.

    Inputs
    Competitor URLs · Focus
    Tools
    fetch_urlweb_search
    Output
    Document + table

  • Data extraction

    Turns files or pages into a clean table, with a source reference for every single cell.

    Inputs
    Files or URLs · Columns, or let it propose
    Tools
    read_filefetch_urlcreate_table
    Output
    Table

  • Draft email

    Writes an email or message from your context and goal, with variants. Sending waits for you.

    Inputs
    Recipient · Goal · Tone · Prior thread
    Tools
    read_filefetch_url
    Output
    Email draftApproval required to send

  • Structured extraction

    Pulls contacts, invoices, events or products into JSON that matches a schema, with a validation report.

    Inputs
    File or URL · JSON schema or preset
    Tools
    read_filefetch_url
    Output
    JSON + table

  • Custom task

    Pro

    Your own instructions, inputs and tool selection, with an approval policy you choose.

    Inputs
    Instructions · Any inputs
    Tools
    Any allowed tool
    Output
    Document, table or JSON

[05]Tools & integrations

Safe tools run. Side effects wait for you.

Agents only touch the tools you allow, per task type and per board. Reading is immediate; anything that sends, posts or writes elsewhere asks first.

Read the API reference
Safe · runs immediatelySide effect · asks for approvalPlatform · Pro
  • Runs
    web_searchTitles, URLs and snippets
  • Runs
    fetch_urlReadable text, robots respected
  • Runs
    read_filePDF, DOCX, XLSX, CSV, images
  • Runs
    create_tableTyped columns, cell sources
  • Runs
    generate_documentMarkdown with citations
  • Asks first
    send_emailGmail, Outlook or SMTP
  • Asks first
    post_to_slackChannels you choose
  • Asks first
    create_notion_pageInto a page or database
  • Asks first
    upload_to_driveInto a folder you pick
  • Asks first
    call_webhookSigned JSON payloads
  • Pro
    Custom HTTP toolsYour APIs, no code
  • Pro
    REST API & webhooksCreate cards, read runs

[06]By design

Built to be trusted with real work.

No borrowed logos. These are guarantees the product keeps on every run, whoever is watching.

Typical run

< 90 s

A standard research brief, from Running to Review.

Unapproved side effects

0

Emails, posts and webhooks always wait for a person.

Work in the open
Free in the demo

3 runs

No sign-up, no card. The workspace lasts 30 minutes.

On the record

100 %

Every plan, tool call, source and cent, for every run.

[07]Pricing

Simple, clear plans.

Start free with 20 runs a month. Pro is 29 euros per workspace, for teams that delegate every day.

Compare every feature

Free

For trying agents on real work.

€0/ forever

No card needed. One workspace, one board.

Start free€0

Includes

  • 20 runs per month
  • 1 concurrent run
  • 1 board, up to 2 members
  • 5 built-in task types
  • Standard model tier
  • Safe built-in tools
  • Exports: Markdown, CSV, JSON
  • 30-day history · Board and List views
Most popular

Pro

For teams that delegate every day.

€29/ month

Per workspace, 5 seats included · +€9 per extra seat

Choose Pro€29/mo

Everything in Free, plus

  • Unlimited runs (fair use: 2,000 / month)
  • 5 concurrent runs
  • Unlimited boards and members (5 seats included, +€9 per seat)
  • Custom task type
  • Standard and Advanced model tiers
  • Side-effecting and custom HTTP tools
  • Slack, Notion, Drive, Email, webhooks and API
  • Card templates and recurring schedules
  • DOCX, PDF and XLSX exports
  • Unlimited history · Timeline view · Run comparison
  • Artifact library · Custom columns
  • Priority support

Custom

Built with Asrar

Need agents that work inside your own tools, with your data and your rules? Asrar designs and builds them, the same way we built this product.

Tell us what you need
  • Task types and tools built for your workflow
  • Wired into your own systems and data
  • Approval rules per team and per action
  • EU hosting, or your own cloud
  • SSO, audit exports and a signed DPA
  • An engineer who knows your setup

[08]FAQ

Answers before you need them.

The questions people ask before they hand an agent real work.

Something else? hello@asrar.example. A real person answers.

It reads the card’s inputs, writes a short plan, then works through it with tools: searching the web, reading pages and files, building tables. Every step appears in the card’s live trace, and the run ends with an artifact in Review: a document, a table, JSON or an email draft.

Your backlog is already written. Hand it to an agent.

Open the demo and run three cards in the next two minutes. No sign-up, no card, nothing sent without you.

Try the demoStart free