NewSolo access is available. Invite codes are issued on request.Redeem invite →
The product

BisChat is the enterprise chat workspace on the PrivateStack platform.

Ask questions, draft documents, and search your files in a familiar AI workspace. You choose the model and retention window, and you get a record of every request. Enterprise runs in your own AWS account.

Start with Solo

Solo / Entrepreneur is $49 per month for one seat. Already have an invite code? Enter it at checkout.

Every request logged
Zero-retention inference on hosted tiers
Open-weight model catalog

What your team gets

The chat experience is the easy part. These are the things that make it defensible when someone asks how AI is being used in your organization.

Chat with your team's own documents

Upload the files your team actually works from and group them into knowledge bases, so answers come from your material instead of a general-purpose model's guess. Knowledge base support is included on every plan.

A seat is a named person

Every seat belongs to a real user with a role. Workspace roles come from your console team list: owners and admins administer, members use. Remove someone from the team and their workspace access goes with them.

Every request on the record

Who asked, when, and which model answered, logged for every request and every admin action. Administrators export the audit history on demand, and set the retention window it lives under.

Model choice from the open-weight catalog

Llama, Mistral, DeepSeek, GLM, Qwen, and Gemma. Change the model behind the workspace without changing how your team works or how your application code calls it.

Zero-retention inference

On hosted tiers, inference is processed by a disclosed US-based subprocessor contractually bound to zero retention: prompts are processed in memory, never stored, and never used to train models.

One endpoint for your own code

The same workspace exposes an OpenAI-compatible endpoint. Point existing LangChain, LlamaIndex, or SDK calls at it, and those requests are logged the same way chat sessions are.

In practice

What using it actually looks like

No training program, no prompt-engineering course. If someone on your team has used an AI chat app before, they already know how to use this one.

  • Ask a question in plain language and get an answer from the model your administrator selected.
  • Point a conversation at a knowledge base so it answers from your documents.
  • Switch models mid-project when a different one fits the task better.
  • Work in a shared team workspace instead of everyone opening a personal AI account.
  • Let administrators see usage per seat, per model, per day.
  • Have every one of those actions land in an audit trail you can export.

What an administrator controls

Model available to the team
open-weight catalog
Seats and roles
admin / member
Knowledge bases
your documents
Audit retention window
you set it

Administrators also export the audit history on demand, so a request for evidence does not turn into an engineering project.

Security and control

Two ways to hold the control

On our hosted tiers, your control is contractual and the data path is disclosed. On Enterprise, your control is architectural, because the workspace runs in the AWS account you already own. Those are different guarantees, so we describe them separately instead of blending them.

Solo and Team

Control by contract and disclosure

We host the workspace. Your control comes from written terms, a disclosed data path, and settings your administrators own.

  • An isolated per-tenant workspace with its own data store and scoped credentials. No shared workspace content between tenants.
  • Inference is processed by a disclosed US-based subprocessor contractually bound to zero retention. Subprocessor names are provided in our DPA and under NDA.
  • Encrypted in transit and at rest, across every connection between users, the workspace, and inference.
  • Audit history is retained on the window your administrators configure, and is exportable at any time.
  • We do not use your prompts, responses, or knowledge-base content to train models, and our inference subprocessor is contractually barred from it too.

Enterprise (AWS BYOC)

Control by architecture

We deploy the workspace inside the cloud account you already own, so the guarantee is structural rather than contractual.

  • BisChat is deployed into your own AWS account through an assumed, least-privilege IAM role.
  • Prompts, outputs, and knowledge bases never leave your VPC.
  • Audit logs and governance records stay inside the account you control, under your own retention and access policies.
  • DSE engineers deploy and operate the stack alongside your security team, including architecture review and a data-flow walkthrough.
  • Enterprise BYOC targets AWS today. Other cloud environments are scoped case by case through sales.

What is not finished yet

  • SOC 2 Type II: certification in progress, not complete.
  • SSO with your own identity provider (Okta, Google, Azure AD): on our roadmap, not shipped. Users sign in to their workspace today.
  • Enterprise BYOC targets AWS. Other clouds are scoped case by case through sales.

The full picture, including subprocessors and the data flow, is on the Trust Center and the security architecture page. Neither is gated.

At a glance

  • BisChat is the private AI chat workspace inside PrivateStack: the everyday chat window your team opens to ask questions, draft, and search your own documents, with the models, the retention, and the audit trail under your control. It is a familiar AI chat interface for your team, on a stack your organization controls.
  • Your team chats, drafts, and asks questions of their own documents through knowledge bases. Your administrators choose the model, the roles, and the retention.
  • Models: open-weight families including Llama, Mistral, DeepSeek, GLM, Qwen, and Gemma, swappable without changing application code.
  • Every request is logged (who asked, when, and which model answered) and the audit history is exportable by your administrators.
  • Hosted-tier inference is processed by a disclosed US-based subprocessor under contractual zero-retention terms: prompts are never stored and never used for training.
  • On Enterprise, BisChat is deployed inside your own AWS account, so prompts, outputs, and knowledge bases never leave your VPC.
  • Solo / Entrepreneur is $49 per month for one seat, by invitation. Team is $99 per user per month for up to 9 seats, sales-assisted. Enterprise is scoped through sales.
  • Built by Data Science & Engineering Experts, Inc. SOC 2 Type II certification in progress; SSO with your own identity provider is on the roadmap.

Put your team on a workspace you control

Solo / Entrepreneur is $49 per month for one seat and you can sign up directly. Team is $99 per user per month for up to 9 seats, and Enterprise is scoped to your infrastructure. Both go through us.

Sign up for Solo

Already have an invite code? Enter it at Solo checkout. No trial, no token-metered billing.