Integrations · Agents
See Langfuse cost with all your other spend
Turn the Langfuse traces you already collect into agent cost by user, session and step, next to the rest of your AI and cloud spend.

Langfuse cost and usage
Turning Langfuse traces into cost means keeping model prices current
Langfuse records what your agents do. Turning that into cost means maintaining model prices, and the result still sits apart from your cloud and AI bills.
- Model prices kept by handCost calculated from normalized model and usage data
- Traces without a cost viewAgent cost calculated from your existing Langfuse telemetry
- Sessions read trace by traceSession and step-level view of agent activity
- No user viewAttribution by user and session
What you can do
See agent cost from your Langfuse traces
Group by agent, model, user, session, step or tool
- GPT-4o$4.1K
- Claude Sonnet 4$3.8K
- Other$1.8K
See everything
Agent cost from traces
See agent cost calculated from the Langfuse telemetry you already collect, with no price setup in Langfuse.
Sessions and steps
Follow agent activity by session and step, including tools.
Users and sessions
Attribute cost by user and session.
How it works
Connect Langfuse with your project keys
No Mavvrik SDK needed. Mavvrik does not collect prompt or response text. Mavvrik holds a current SOC 2 Type 2 report and supports single sign-on and role-based access.
1 Connect
Add your keys
Enter your Langfuse public key, secret key and base URL, for Langfuse Cloud or self-hosted. Add an agent ID key to your trace metadata.
2 Data flows in
Mavvrik polls your traces
Every 10 minutes, Mavvrik reads observations and traces: model, tokens, latency, status, user, session and agent metadata.
3 See your costs
Connect discovered agents
Mavvrik lists the agents it finds in your traces, and you connect them. The first collection looks back 24 hours.
Total cost
Langfuse is just one part of your total AI cost
Some of the other costs you may want to track alongside Langfuse
AgentsLangChainSee cost per agent and workflow, traced to each model call.View integration →
AI platformsOpenAISee OpenAI API cost and token usage by model and operation.View integration →
AI toolsClaude Code and CoworkSee Claude Code and Cowork spend by user and session.View integration →
CloudAWSSee AWS cost by service, account and team, including Bedrock.View integration →
FAQ
Questions you might have
Do we need to set model prices in Langfuse?
No. Mavvrik calculates cost from normalized model and usage data.
Does Mavvrik load older traces?
No. The first collection looks back 24 hours, and older traces are not backfilled.
Does it work with self-hosted Langfuse?
Yes, over HTTPS.