Straight answers.

Brrainy is a memory layer that works across every AI assistant you use. It captures what you work on, strips the noise, organises it into your own space, and gives it back when it is relevant - so you stop re-explaining yourself to every new chat.

One memory, read by every assistant. The browser extension captures conversations automatically from ChatGPT, Claude, Gemini, Perplexity, DeepSeek and Grok, and because the MCP server is hosted over HTTP, any MCP-capable assistant or agent connects to that same memory. Switching models stops costing you your context, because the context never lived inside the model.

Spaces, routing rules, and role-based access. A space keeps one body of work separate - personal stays personal even when you join an organisation.

Brain plugins you install from the marketplace - briefs, open loops, patterns, anomalies and knowledge gaps among them. Access follows your plan tier.

ChatGPT's memory works inside ChatGPT. It does not travel to Claude or Gemini, and you cannot take it with you. Built-in memory is free and needs no setup, which Brrainy does not match - you connect your sources first. What you get for that is one memory every assistant reads, that stays yours when you change tools.

Rewind's capture was discontinued after Meta acquired Limitless in December 2025. Brrainy is a different shape: it captures from your AI assistants and your connected sources rather than continuously recording your screen, and the memory is exportable and yours.

Free is $0. Pro is $20 a month. Team is $15 per member per month with no seat minimum. Enterprise is custom-quoted and adds SSO with OIDC and SAML plus audit log export. You can bring your own model key on any plan, including Free.

Yes. Free is a real tier, not a trial - capture, search, consolidation, briefs, and reminders all work within quotas, and your own model key works on it. Paid tiers raise the caps and add deeper analysis, collaboration, and governance.

The gate holds rather than quietly degrading. Limits come from your plan's entitlements and are always enforced, so you see the ceiling instead of wondering why results got worse. Adding your own model key removes the AI-spend ceiling on every tier, including Free.

Because of what the price includes. Tools in this category typically meter the AI work itself - you pay per memory stored and per search run. Bring your own key and those operations cost you provider rates and nothing to us, across chat, embeddings, reranking, OCR, and speech-to-text. The headline price is similar; the bill after a heavy month is not.

Plans are monthly, and bring-your-own-key means the compute side was never ours to lock up - your key points at your provider, and a custom endpoint can point at a model you run yourself. Billing is Stripe for self-serve; Enterprise is custom-quoted.

Yes, on every plan including Free. Bring-your-own-key is honoured across all five paid operations - chat, embeddings, reranking, OCR and speech-to-text - not chat alone. Indexing a large corpus is the expensive part, and that is the part most tools keep billing for.

No - Brrainy is a hosted service, and we do not sell or support a self-hosted edition. What you do control is the AI: bring your own model key and chat, embeddings, reranking, OCR and speech-to-text all run on your provider account, on every plan including Free. A custom endpoint can point at a model you run yourself.

Chat, embeddings, reranking, OCR, and speech-to-text - all five, not just chat. Add a key in Settings and Brrainy resolves it for that space instead of debiting your wallet, so processing keeps running even at a zero credit balance. Indexing a large corpus is the expensive part, and that is exactly the part your own key covers.

Yes. Point the custom endpoint at any OpenAI-compatible runtime - Ollama, llama.cpp, or vLLM - and it resolves the same way as a hosted provider key.

Yes, over MCP. There are exactly two tools, brain_recall and brain_remember, so an agent can look something up or write something back without a sprawling API surface. Humans and agents read the same space-scoped memory.

Connectors for mail, calendar, notes, chat, issues and documents (Gmail, Google and Outlook calendars, Notion, Obsidian, Apple Notes, Slack, Discord, GitHub, Linear, Jira, Google Docs and more), each a real per-provider implementation rather than a stub. Automatic browser capture covers ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek, plus ambient capture on the desktop. Beyond those, any MCP-capable assistant or agent can read and write the same memory over the hosted MCP server.

Yes. Team is $15 per member per month with no seat minimum, for up to 50 members. Everyone works in shared spaces with role-based access, and your agents read the same organisation memory over MCP - so context survives people leaving.

Team is $15 per member per month for up to 50 members, with no seat minimum - a team of four can buy it. Everyone shares spaces and one memory with role-based access, and your agents read the same organisation memory over MCP. Above 50 members it becomes Enterprise.

SSO with OIDC and SAML, plus audit log export. Talk to us about your identity provider - Enterprise is quoted per organisation, so the terms are set with you rather than picked off a page.