Technical FAQ

Brainverse agent teams integrate with your existing stack through three primary mechanisms:

• Claude API (Anthropic) — the underlying model layer powering agent reasoning and generation. No separate model infrastructure required on your end.
• MCP (Model Context Protocol) — the integration layer that lets agents securely communicate with external systems. We configure MCP connectors for each tool in your stack (CRMs, document stores, databases, SaaS platforms) during the deployment phase.
• Direct API integrations — for tools with well-defined APIs (GoHighLevel, Google Workspace, etc.), agents call those APIs directly. Credentials are provisioned with least-privilege access.

The specific integrations required are defined during the discovery call, based on your actual workflows — not a generic list.

Minimal. Brainverse deploys on our own infrastructure (Vercel, Supabase, Railway). You do not need to provision servers, manage containers, or maintain any agent runtime. What you do need: API credentials for the tools your agents will connect to (CRM, document storage, etc.), and access to your BrainSync dashboard for monitoring and approvals. Everything else is managed on the Brainverse side.

Only what is explicitly needed for the workflows we deploy. Data access requirements are scoped during the discovery phase — you will receive a clear list of what each agent can read and write before deployment begins. Agents do not receive broad access to your systems. Each integration is configured with the minimum permissions needed for its specific function.

No. Agents do not establish direct connections to your internal databases unless that is explicitly agreed upon and configured as part of the deployment. The standard integration pattern uses your existing SaaS APIs and document stores — tools you already have access controls on. If direct database access is required for a specific workflow, we discuss the architecture and security requirements before proceeding.

Data you share with Brainverse — documents, records, workflow inputs — is:

• Used only to perform the specific tasks you have engaged us for
• Not retained beyond operational necessity
• Not used to train AI models (covered by enterprise agreements with our model providers)
• Not shared with other clients or third parties for their purposes

See the Security Posture page for full details on encryption and access controls.

Yes. The BrainSync dashboard provides full visibility into agent activity — every action is logged, every session is tracked. You can see what each agent did, when, and what the outcome was. This is not an audit log buried in a settings page — it is the primary interface for understanding agent activity across your deployment.

Agents communicate with external systems through two mechanisms:

• MCP (Model Context Protocol) — a structured protocol for agent-to-system communication. Each integration is defined as an MCP server with explicit, scoped capabilities. Agents cannot perform actions outside what the MCP server allows.
• REST API calls — for tools with standard REST APIs, agents make direct HTTP calls using pre-provisioned credentials. All calls go through standard HTTPS.

Agents do not have shell access to your systems, cannot install software, and cannot execute arbitrary commands on your infrastructure.

Each client deployment runs as an isolated agent team with its own configuration, memory, and credentials. Data is separated at the database layer using Row-Level Security (RLS) — one client's data cannot be read by another client's agents, even within the same database infrastructure. There is no shared runtime state between deployments.

Agent workflows are designed with human-in-the-loop checkpoints for actions that have significant downstream effects. Agents log every action to the BrainSync dashboard, so any errors are visible and attributable. For irreversible actions (sending emails, modifying records), standard practice is to configure an approval step so a team member reviews before execution. The specific guardrails for each workflow are defined during deployment based on your risk tolerance.

Typically 4 to 6 weeks from the discovery call to a fully operational agent team. The phases:

• Discovery (Week 1) — map your workflows, identify integration points, define agent scope and data access requirements
• Build (Weeks 2–4) — configure agents, set up integrations, establish the BrainSync dashboard
• Testing and handoff (Weeks 5–6) — validate agent behavior against real workflows, train your team on the dashboard, go live

Timelines vary based on the number of integrations and the complexity of the workflows involved.

The ongoing support model includes:

• Daily improvement cycles — the agent team runs self-improvement analysis each night. Surfaced improvements go through an approval workflow before any changes are deployed.
• BrainSync dashboard — real-time visibility into agent activity, session logs, and performance metrics
• Weekly proof-of-life reports — a summary of what the agents accomplished, flagged exceptions, and recommended actions
• Direct access — Jeff Leggett is the point of contact for all client accounts. No support ticket system.

Yes. The agent team is designed to grow with your business. New agents can be added to existing deployments as you identify new workflows to automate. The BrainSync dashboard tracks all agents and their activity, so adding to the team does not create visibility gaps.