Agents & MCP

Agents that act, with guardrails you control

A Branja.ai agent is a model, a set of tools exposed over MCP, memory, and rules about what it may do without asking.

Connect

Attach MCP servers or custom connectors. Each tool declares its inputs and permissions.

Define

Write the goal, the tools it may use and the approvals it needs before sending or spending.

Run

Trigger on a schedule, an event or a message. Long jobs resume if a step fails.

Review

Every decision, tool call and output is traced, scored and replayable.

Built on open protocols

We use the Model Context Protocol so the same connector works with any compatible model or client. You are never locked into one vendor's plugin store.

Hosted MCP servers for common SaaS tools
A connector SDK for your internal APIs
Per-tool scopes, rate limits and audit logs

{
  "mcpServers": {
    "branja-hubspot": {
      "url": "https://mcp.branja.ai/hubspot",
      "scopes": ["contacts:read", "deals:read"]
    },
    "branja-drive": {
      "url": "https://mcp.branja.ai/drive",
      "scopes": ["files:read"]
    }
  },
  "approvals": ["email.send", "payment.create"]
}

Want an agent for a specific workflow?

Tell us the workflow and the systems involved. We will scope a first agent and a pilot.

Scope an agent