Bring Your Own Agent

Connect any agent straight to your data.

Claude, ChatGPT, Gemini, or any MCP client connects to your data through a single authenticated endpoint — reading the same governed metrics Fi does, not raw tables it has to guess at.

One endpoint, any agent

Your data, exposed through MCP.

Definite speaks the Model Context Protocol, so any MCP-aware agent can connect to your lakehouse and semantic layer through one authenticated, governed port. The agent you already use becomes a data analyst — grounded in your real definitions.

Your agents
Claude
ChatGPT
Gemini
… or any MCP client
───►
mcp://definite
authed · governed
───►
Your data
Lakehouse
Semantic layer
same metrics Fi uses

The agents your team already uses, grounded in your data.

Any MCP client

Claude, ChatGPT, Gemini, or your own agent — anything that speaks MCP connects without custom integration work.

Authenticated & governed

Access runs through one authed endpoint with the same permissions and definitions as the rest of the platform.

The same semantic layer

External agents query governed metrics — so their answers match your dashboards and Fi, instead of guessing against raw tables.

Beyond read-only

Agents can explore, query, and act on your data through the tools the protocol exposes — not just fetch rows.

Connect an agent in three steps.

  1. 1Point it at the endpointAdd Definite's MCP endpoint to your agent of choice — no bespoke connector to build.
  2. 2AuthenticateThe agent connects with scoped, governed access to your data.
  3. 3Ask awayYour agent now answers from your governed metrics, consistent with everything else in Definite.
§ Metrics as code

Dashboards you can diff.

Your data stack lives in your repo, not a vendor's UI. Code-review KPIs. Deploy dashboards like you deploy code. The agent opens the PR; you approve the metric.

models/sales.yaml● on main
name: deals
sql_table: lake.hubspot.deals

measures:
  - name: total_revenue
    description: Total closed revenue in USD
    sql: amount
    type: sum

  - name: win_rate
    description: Percentage of deals marked as won
    sql: "COUNT(CASE WHEN status = 'won' THEN 1 END)::float / COUNT(*)"
    type: number

dimensions:
  - name: stage
    description: Current pipeline stage
    sql: stage
    type: string

  - name: closed_date
    description: Date the deal was closed
    sql: closed_at
    type: time
§ Agent, meet governance

Fast to build.
Safe to trust.

The agent writes the queries. The platform keeps them honest — so your CFO can trust the numbers that land on their dashboard.

01
Semantic layer, not raw tables

Agents query defined metrics — consistent numbers across every dashboard, every time.

02
Row-level security, by default

Define who sees what once. Every query, export, and dashboard respects the rules.

03
Full audit trail

Every change tracked. Every query logged. Your CFO can trust what the agent built.

More from Agents & Platform

Your answer engine
is one afternoon away.

Book a 30-minute call. We'll build your first dashboard on the call — or you can stop paying us.