Pricing

Simple plans. Serious deliberation.

Every plan runs on your own machine, Linux or macOS. Cancel anytime through our merchant of record, Polar.

Basic

For individuals getting started
$19/mo
  • 3 AI participants
  • Sequential sessions
  • Roundtable mode
  • 3 file references / 2 outputs
Buy Basic

Pro Max

For power users & builders
$79/mo
  • 6 AI participants
  • Everything in Pro
  • 20 file references / 16 outputs
  • SDK included (C API + Python)
Buy Pro Max

Enterprise

For teams and organizations
Contact us
  • Up to 16 AI participants
  • Unlimited file refs & outputs
  • Server mode
  • Seat licensing & user management
Talk to sales

SDK Add on

For Basic and Pro subscribers

$15/mo
  • C API + Python SDK
  • Run deliberations from your own code
  • Requires an active Basic or Pro plan
  • Already included in Pro Max & Enterprise
Buy SDK Add on

Compare plans

FeatureBasicProPro MaxEnterprise
AI participants per session34616
Sequential sessions
Parallel sessions
Roundtable mode
Directed mode
File references per session31020Unlimited
File outputs per session2816Unlimited
File exchange between models: text + source code
File exchange between models: PDF
File exchange between models: images, audio, any binary
Cloud + local models (Ollama or any OpenAI compatible server)
Web GUI
Command line interface
Roundtable rounds per session (max)82040100
Total turns per session (max)401203001000
SDK (C API + Python)Add-onAdd-on
Server mode
Seat licensing & user management
In the box

In every package: nine demos.

Whichever plan you choose, nine worked sessions come in the box, each with its own guide. Load one, add your keys, press start.

Legal strategy

The flagship: structured multi-model deliberation over a real case brief, ending in a strategy memo.

Adversarial code review

Review by deliberation instead of by linter: models defend and attack the same code.

Code co-writing

Models that genuinely disagree before they build: a rate limiter argued into existence.

Two painters

Binary exchange, live: one model paints an image, another repaints over that actual file.

Incident war room

Directed mode: the moderator routes every turn to the right specialist, like a real incident commander.

Guided intelligence

A reasoning model mentors a non-reasoning one, and the pair beats either alone.

Independent estimates

Parallel transport: everyone answers simultaneously, so no estimate anchors the others.

Moderator judgment

The panel runs until the moderator judges the work is done, not until a counter says so.

Hello session

The smallest working session: a moderator and one participant, your first two minutes with MCS.

The terms, in plain words

Turn and round

A turn is one participant speaking once. A round is one full pass through the panel: the moderator opens, each participant speaks, and the moderator closes.

Roundtable and directed

These are the two modes, and they decide how turns are ordered. In roundtable, participants speak in a fixed order, round by round. In directed, the moderator chooses who speaks next, turn by turn. Directed mode always runs sequentially.

Sequential and parallel

This is how the panel answers within a round. Sequential means one at a time, each seeing what came before. Parallel means the panel answers at the same time.

File references and outputs

File references are files you give the panel to read. File outputs are files the panel produces during the session and hands to each other and to you.

When the stakes are high, one AI model may not be enough

Every plan includes

🔑

A real license, not a login

Your license key works on your machine, with thirty days of offline operation between check ins. MCS keeps working when your network doesn't.

⚙️

The engine and the web GUI

A local web interface on every plan, plus a command line on Pro Max and Enterprise. Sessions are yours: inputs, outputs, logs, all on your disk.

📚

Documentation and demos

The user guides are on the web, see the Documentation. Ready to run demo sessions are included in every download.