Teams and agents

Use the AI subscriptions you already pay for

Bring your own model means Brainwrite starts the agent CLIs already on your Mac, such as Claude, Codex, Grok, Cursor, Kimi, Factory Droid, Antigravity, OpenCode, Qwen, Hermes and Pi, and shows them in one consistent chat. It also supports the Mistral API and local models. Your provider bills your usage, and Brainwrite does not route your messages through its own model account.

Download for Mac Read the docs

The right engine for the job

Review the checkout code in this repository and list the three riskiest functions. Do not change files yet.

Your Chief of Staff put a team on it

Nova· Chief of Staff
Dash· Triage bot
Atlas· Rigel, QA and Release Engineer

1. applyDiscount: no test covers stacked coupons.
2. chargeCard: retries without an idempotency key.
3. formatTotal: rounds before tax instead of after.
Running on your own Codex login; the research bot stays on Claude.

DoneDone. Nothing was sent.

An illustrated example. Your team works on your own files and apps.

Split scene: at 11:04 pm a developer rubs his eyes alone at his desk; at 7:04 am he drinks coffee while AI agents Atlas and Dash show finished work.

See it in the app

Bring your own model in Brainwrite.

A bot's model picker open in Brainwrite: Claude models listed, with Claude, Grok and Cursor in the engine rail and an Add API keys link.

Why it matters

On your own, with a chatbot, and with Brainwrite.

	Doing it yourself	A single chatbot	Brainwrite
Which models	One app per provider	The models that product offers	Claude, Codex, Grok, Cursor, local models and more, per bot and per thread
Who bills you	Each subscription separately	That product's plan	Your own login or API key; your provider bills usage
Mixing models	Copy work between apps	Usually one provider's models	A Codex coding bot and a Claude research bot, at the same time
Private or offline work	Run a local model by hand	Usually needs the internet	Local models through OpenCode, Ollama or LM Studio

How it works

How Bring your own model works.

01
Install and sign in to a CLI
Install a supported agent CLI and finish its own login, or keep the one you already use. Restart Brainwrite after installing so it is detected.
02
Check detection
Open Settings → Engines. If an engine is installed somewhere unusual, or behind a version manager or wrapper, set its executable path there.
03
Pick a model per bot
In a bot's model picker, choose the engine and model. The picker groups models by provider, marks defaults, and separates cloud engines from local ones.
04
Override per thread
Each thread remembers its own model, so one bot can use a fast model for triage and a stronger one for a hard problem.

Good to know

The details.

Mistral without a CLI

Save a Mistral API key in Settings → Connections and choose Mistral (API) in the picker. Test checks catalog access without sending a chat message. This connection does not get filesystem access.

Local models

Run local models through OpenCode, Ollama or LM Studio for private or offline work. Brainwrite itself keeps working offline for up to 7 days since its last account check.

Antigravity with your Google account

Choose Install official Antigravity, then Sign in with Google. Brainwrite checks the download from dl.google.com against a pinned size and SHA-256 and never asks for your Google password.

Live model lists

Supported engines refresh their catalog from the installed CLI. Codex reads the signed-in CLI's live list, and Brainwrite offers an update command when a newer stable Codex release exists.

One broken engine does not break the rest

An engine that is not available stays visible with a reason, instead of taking the other engines down with it.

The edges, plainly

AI usage is not included in any Brainwrite plan. Your CLI login or API key decides what you can use and who bills you.
One subscription does not give access to another provider's models.
If an engine returns a 401 or asks you to sign in, finish that CLI's own login, then refresh detection in Settings.

Full reference in the documentation

Use cases

Jobs this feature does for you.

Works with

Apps it's often used with.

FAQ

Questions about Bring your own model

Which AI models can I use in Brainwrite?

Brainwrite starts the agent CLIs you already have: Claude, Codex, Grok, Cursor, Kimi, Factory Droid, Antigravity (Google), OpenCode, Qwen, Hermes and Pi. It also connects to the Mistral API with your key, and to local models through OpenCode, Ollama or LM Studio. You pick the engine and model per bot.

Does Brainwrite include AI usage in its price?

No. Neither the Brainwrite Cloud plan nor the Organisation plan includes AI usage. Brainwrite starts the agent CLIs you sign in to yourself, or uses your API key, and your provider bills that usage under its own terms and limits.

Can I run AI agents on local models?

Yes. Local models run through OpenCode, Ollama or LM Studio, and the model picker shows local engines separately from cloud-backed ones. A local-model bot suits private or offline work. Brainwrite itself keeps working for up to 7 days without reaching its account service.

Can different bots use different AI models at the same time?

Yes. Pick an engine and model per bot, and each thread can keep its own model too. A coding bot can run on Codex while a research bot runs on Claude and a private bot runs on a local model, all at once.

Works well with