Integrations · Agents

See LangChain cost with all your other spend

Track cost per LangChain agent and workflow, traced down to each model call, next to the rest of your AI and cloud spend.

LangChain
Mavvrik
VisibilityCost per agent and workflow
TracesFramework steps and model calls in one trace
AttributionCost by customer, user and use case

LangChain cost and usage

Agent cost is spread across steps, models and tools

One LangChain run can call several models and tools. Model provider bills show totals by model, not which agent, customer or workflow ran up the cost.

Without MavvrikWith Mavvrik
  • Provider bills by model onlyCost per agent and workflow
  • Steps and model calls traced separatelyFramework activity and model calls in one execution trace
  • No customer or use case viewAttribution by customer, user, session and use case
  • Non-LLM costs left outMetered non-LLM usage in the same workflow cost
  • List prices onlyCustom LLM pricing for negotiated rates

What you can do

See and attribute LangChain agent cost

LangChain agentsOctober
$18.3K+11.5% vs. Septemberacross 9 agents
support-triage
$6.2K
contract-review
$4.8K
research-assistant
$3.9K
onboarding-bot
$2.1K
All other
$1.3K

Group by agent, provider, model, session, user, customer or use case

By provider
  • OpenAI$9.6K
  • Anthropic$6.9K
  • Other$1.8K
By use case
  • Support$8.0K
  • Legal review$4.8K
  • Research$3.9K
  • Onboarding$1.6K

See everything

  • Cost per agent and workflow

    See what each agent and workflow costs to run.

  • Execution traces

    See framework steps and the model calls beneath them in one trace, with tokens, duration and status.

  • Customer and use case

    Attribute cost by customer, user, session and use case.

Allocate accurately

  • Non-LLM usage

    Meter usage outside the model, such as OCR pages or API calls, into the same workflow cost.

  • Negotiated rates

    Apply custom LLM pricing for the rates you negotiated.

How it works

Instrument LangChain with the Mavvrik SDK

Mavvrik captures agent steps and the model calls beneath them. Prompt and response text is not sent by default. Mavvrik holds a current SOC 2 Type 2 report and supports single sign-on and role-based access.

1 Connect

Add the SDK

Register the agent in Mavvrik, install the Mavvrik SDK with LangChain instrumentation for Python or JavaScript and TypeScript, and call instrument() before importing LangChain. LangGraph is supported.

2 Data flows in

Spans flow to Mavvrik

Each agent step and model call, with provider, model, input, output, cached and reasoning tokens, duration and status.

3 See your costs

Activity within 10 minutes

Processed activity shows up in Mavvrik within 10 minutes.

Total cost

LangChain is just one part of your total AI cost

Some of the other costs you may want to track alongside LangChain

See all integrations →

FAQ

Questions you might have

Does Mavvrik receive prompts or responses?

Not by default. The SDK sends model, token, duration and status data.

Is LangGraph supported?

Yes.

How soon does data show up?

Within 10 minutes.