Multi-Agent
A team of agents for bigger tasks.
The orchestrator assesses the request and, within your plan, picks leads and workers. You see the scenario and the cost of a run before it starts.
New in the iPhone app: your own AI agent with a name, an appearance and a brain that splits work by itself into clones, subagents and agent teams. The guide Clones, subagents and agent teams explains how it works.
Deep Research
320credits
- Plus
- model tokens
- Run time
- 5-15 minutes
- Leads at once
- 4
- Workers per lead
- 3
- Iterations
- 4
- Runs per day
- 80
The team starts only when the task requires it.
You turn on Multi-Agent in the message field. From there the orchestrator picks the team itself, within your plan's limits.
- 01
You turn on Multi-Agent
The switch is in the message field. You write the request the same way as in a regular conversation.
- 02
The orchestrator assesses the task
It handles simpler tasks alone, with tools. Larger ones it splits between leads, and each lead picks workers.
- 03
You know the cost before the start
You see the scenario and the base cost in credits. Model tokens are added after the work is finished.
- 04
The result goes into the conversation
An answer, a table or a report. You can ask follow-up questions right away or request changes.
Four layers of the team.
An orchestrator, seven leads, workers and a critic. Which leads and how many workers the team uses depends on the task and on the plan.
Coordination
Plans, delegates and assembles the result.
It reads the request and decides whether one answer with tools is enough or a team is needed. If so, it splits the work between leads, then assembles their results into one answer. Every Multi-Agent run starts with it.
Leads
Seven specializations. Each lead picks its own workers.
- ZZofiaresearch
Gathers and compares sources, summarizes and checks facts. Leads Deep Research.Web search, news, page reading, conversation history
from Free - WWojtekplanning
Roadmaps, schedules, estimates and priorities. Turns a goal into a plan with stages and deadlines.Plan, calendar
from Base - PPatrykcalendar
Adds dates in a sentence, checks conflicts, and sets reminders and focus blocks in the device calendar.Calendar, reminders
from Base - MMarekanalysis
Calculates, models and detects anomalies: KPI reports, correlations, cleaning up data from spreadsheets and PDFs.Calculator, running code, files
from Plus - BBrunocode
Writes and fixes code, runs it in a sandbox and does reviews: refactors, tests, migrations, data scripts.Python in a sandbox, files
from Plus - AAniacreative
Campaign concepts, names and copy in a set tone. Generates images for the brief.Images, copy
from Plus - WWiktoriavision
Analyzes images, screenshots and photos of documents (OCR), compares them and extracts data.Image analysis, OCR
from Plus
Workers
Each one does a single slice of the task and hands the result to its lead.
- Researchfrom Free
Source review, data synthesis, fact checking, summaries.
- Planningfrom Base
Tasks with estimates and deadlines, calendar conflicts.
- Code and analysisfrom Plus
Tests, reviews, migrations, models and KPI reports. Code runs in a sandbox.
- Contentfrom Plus
Headline variants, copy in a set tone, descriptions and graphics for the brief.
Verification
An additional check in the Deep scenario.
Checks the numbers, consistency and gaps in the team's result. When something does not add up, the task goes back to the lead for another iteration. The critic works in the Deep scenario, available from the Plus plan.
Five scenarios. You know the cost before the start.
The cost of a run in credits, according to the billing catalogue. Model tokens are added on top.
Solo
5credits
The orchestrator answers alone, with tools, without delegation.
Every planMini
20credits
1 lead and up to 2 workers, one iteration. On Free with 1 worker.
Highest on FreeStandard
55credits
2-3 leads and 3-5 workers, two iterations.
From the Base planDeep
140credits
The most agents, a critic and extra iterations.
From the Plus planDeep Research
320credits
Parallel searches and a long report with sources.
5-15 minutes of work| Team limit | Free | Base | Plus | Pro | Max | Enterprise |
|---|---|---|---|---|---|---|
| Leads on the team | 1 | 2 | 4 | 8 | 15 | ∞ |
| Workers per lead | 1 | 2 | 3 | 5 | 8 | ∞ |
| Iterations | 1 | 2 | 4 | 8 | 15 | ∞ |
| Highest scenario | Mini | Standard | Deep | Deep | Deep | Deep |
| Multi-Agent runs / 24 h | 15 | 30 | 80 | 200 | 500 | ∞ |
| Available leads | ||||||
| Zofia (research) | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Wojtek (planning), Patryk (calendar) | none | ✓ | ✓ | ✓ | ✓ | ✓ |
| Marek (analysis), Bruno (code), Ania (creative), Wiktoria (vision) | none | none | ✓ | ✓ | ✓ | ✓ |
Values from the DomaAI billing catalogue. The limits for all features are in the limits table, and model rates are in the pricing in credits.
Questions about Multi-Agent
More in the help centerHow do I know what I'll pay?
Before a run you see the scenario and the base cost in credits. Model tokens are added after the work is finished. If a run fails, the credits return to your account.
How does Deep Research differ from Deep?
Deep is the largest team, with a critic and extra iterations, for any task. Deep Research runs several searches at once and delivers a long report with sources, after 5-15 minutes of work.
How does Multi-Agent differ from automation agents?
Multi-Agent works on the message you just sent and delivers the result in the conversation. Automation agents work on a schedule or via webhook, even when the app is closed. How automation agents work
Do agents have access to my data?
Only to what you give them in the conversation: files, the calendar with your consent, and project memory. We do not train models on your data, and requests go directly to providers.
Turn on Multi-Agent in your next conversation.
Free account, no card. On Free you get the research lead and the Mini scenario, higher plans unlock more leads.
