Get Started

Use this page when connecting an agent to Mavvrik for the first time.

The setup path depends on how the agent already runs. Choose the integration first, complete that connection, then verify one known request before adding optional attribution or metered usage.

1. Choose how the agent connects

Your setup	Start here	Registration model
Python agent application	SDK Integration	Register an SDK agent, then install the Python SDK.
JavaScript / TypeScript agent application	SDK Integration	Register an SDK agent, then install the JavaScript / TypeScript SDK.
eve agent application	Connect eve	Register the eve agent, then configure eve's native OpenTelemetry export. No Mavvrik SDK is required.
Existing Langfuse project	Connect Langfuse	Create a Langfuse connection and connect discovered agents.
LiteLLM proxy	Connect LiteLLM	Create a LiteLLM connection; request metadata identifies individual agents.
Self-hosted n8n	Connect n8n	Register the n8n connection and install the n8n SDK.

If the application uses the Mavvrik SDK, check Supported Agentic Stacks first when provider, framework, vector database, web framework, or runtime compatibility must be confirmed.

2. SDK users: register the agent

Python and JavaScript / TypeScript SDK integrations require an agent identity before SDK installation.

Open Register an Agent and complete the Setup and Connect steps.

You will receive:

MVK_TENANT_ID
MVK_AGENT_ID
MVK_API_KEY

Then continue with:

eve, Langfuse, LiteLLM, and n8n users should follow their integration-specific onboarding pages instead of this SDK registration step.

3. Run one known test

After the integration is connected, generate one known AI request or workflow.

For SDK integrations, use a request with a known session_id or user_id when possible so it is easy to locate.

For eve, Langfuse, LiteLLM, and n8n, use the verification request described in the corresponding integration guide.

4. Verify the first data

Confirm both views:

Home → Agentic → Cost — the expected agent, model usage, and cost appear.
Home → Agentic → Sessions — the execution and captured operations appear.

Allow up to 10 minutes for processed activity to appear unless the integration guide specifies a different collection interval. For eve, new agent data may take up to 24 hours to appear.

Do not continue to optional enrichment until the baseline model usage is visible.

5. Add business context when you need attribution

Use Add Business Context when the same agent serves multiple customers, users, sessions, applications, or business use cases.

Examples:

customer_id = acme
user_id = user-123
session_id = support-456
use_case = customer_support

This lets Mavvrik answer questions such as cost per customer, user, session, application, or use case.

6. Track paid non-LLM usage when needed

Use Track Metered Usage only when the workflow consumes paid resources outside supported model calls.

Examples:

OCR pages;
paid API requests;
document-processing units;
storage or other per-unit services.

A supported model call does not need to be recorded again as metered usage.

7. Configure LLM pricing when needed

Use Configure LLM Pricing when negotiated LLM rates differ from Mavvrik's default model pricing.

This step is optional. It does not affect whether model usage can be captured.

First setup complete

The initial integration is complete when:

the correct integration path is connected;
one known request has completed;
model usage appears in Cost;
execution detail appears in Sessions;
required business context is present;
paid non-LLM usage is added only where applicable;
customer-specific LLM pricing is configured only where applicable.

For missing data, use Troubleshooting Agent Data.