01Drop-in
Keep your provider's API
Change a base URL and add two headers. OpenAI, Anthropic and Gemini APIs work as they are.
Build
Build your AI product on our pipeline.
Privacy, memory, context, masking and routing as services, for teams whose products handle personal data. You build the product; the pipeline does the rest.
In development. Not available yet.
What you get
Everything between your product and the model.
02Your customers
Each one kept apart
Onboard your customers with their own tenants, plan and policy. A search never crosses between them.
03Interfaces
Your stack, your language
A command-line tool, TypeScript and Python SDKs, an MCP server for agents and an HTTP API.
04Policy
Your rules, signed
Providers, models, regions, budgets and masked-only traffic. A policy can only narrow.
05Keys
Your keys, your endpoints
Bring your own model keys and private model endpoints.
06Hosts
On your own hosts
Brain Ways and Brain Authority are built to run on private hosts.
Who does what
You build what makes your product yours.
You build
- Your product and its interface
- Your prompts, workflows and agents
- Which models, regions and budgets your customers get
The pipeline handles
- Identity, signed policy and audit
- Memory with provenance, and context in scope
- Masking before any model
- Routing and failover, with no content kept
For example
Products that need personal data, and must keep it in scope.
Assistants
Answers over a customer's own documents, with sources.
Support
Replies that know the case history, without sending names to a model.
Agents
Agents that read and add to memory through the MCP server, in scope.
Regulated teams
Search and drafting for health, finance and legal work.
From the terminal
Set it all up from the command line. Then script it.
The CLI runs every operation: projects, Spaces, keys, policy, memory and search, with structured output a script can rely on.