--- title: "Connect Langfuse to Nvidia - APIANT.ai" description: "When \"New Trace\" fires in Langfuse, your agent runs \"Rerank documents\" in Nvidia." canonical: "https://apiant.ai/connect/langfuse-to-nvidia" --- # Connect Langfuse to Nvidia When **New Trace** fires in [Langfuse](https://apiant.ai/connections/langfuse), your agent runs **Rerank documents** in [Nvidia](https://apiant.ai/connections/nvidia). Describe the whole flow and it is built, tested on every scenario and kept running. Langfuse **1** trigger Langfuse **14** actions Nvidia **5** actions What the catalog holds today. Ask for anything missing and your agent reads the API documentation and adds it. ## What do you want Langfuse and Nvidia to do together? Give your AI client this address. `https://mcp.apiant.ai` One connection does both: ask your client to act in an app once, or have it build an automation that keeps running. Any client that takes a remote MCP server. Nothing to install. Create your free account, then sign in once in your browser. ### Set it up in your client ### Claude Code Run this in the directory you want to work in. `claude mcp add --transport http -s project apiant-ai https://mcp.apiant.ai` Start Claude Code there, approve the server, and sign in in the browser page it opens. [Claude Code guide (opens in a new tab)](https://apiant.ai/docs/mcp/claude-code) ### Claude, web and desktop Open Customize, then Connectors, then Add custom connector, and paste the address. On Team and Enterprise an owner adds it first. [Setup guide (opens in a new tab)](https://apiant.ai/docs/mcp) ### Cursor Add the address to your MCP configuration as a server URL. Cursor signs you in in your browser. [Setup guide (opens in a new tab)](https://apiant.ai/docs/mcp) ### Zed Open Settings, then AI, then MCP Servers, then Add Server, then Add Remote Server. [Setup guide (opens in a new tab)](https://apiant.ai/docs/mcp) ### Codex and the ChatGPT desktop app Open Settings, then MCP servers, then Add server, and choose Streamable HTTP. From the terminal, Codex takes the address with its own add command and then signs you in. [Setup guide (opens in a new tab)](https://apiant.ai/docs/mcp) ### Any other MCP client Paste the address where the client asks for a remote MCP server URL, and name the server apiant-ai where it lets you. APIANT's skills reach the client over the same connection. [Setup guide (opens in a new tab)](https://apiant.ai/docs/mcp) Example flows ## What Langfuse can start in Nvidia - When **New Trace** in Langfuse then **Rerank documents** in Nvidia - When **New Trace** in Langfuse then **Create chat completion** in Nvidia Starting points drawn from the triggers and actions in the catalog today. Describe the flow you need, at any depth, and your agent builds it and tests every scenario before it ships. Everything your agent can do with Langfuse and Nvidia ## Triggers and actions 1 trigger ยท 19 actions - Trigger **New Trace** Polling trigger: fires for newly created traces (dedup by trace id / timestamp). - Action **Add Comment** Add a comment to a trace/observation/session/prompt object. - Action **Add Dataset** Create a new dataset. - Action **Add Dataset Item** Create an item (input/expected_output/metadata) in a dataset. - Action **Add Prompt** Create a new prompt version (text or chat). - Action **Add Score** Create a score attached to a trace/observation/session (numeric, categorical, or boolean). - Action **Delete Trace** Delete a trace by id. - Action **Get Project** Reads the Langfuse project that the connection's project-scoped API key belongs to. - Action **Get Prompt** Get a prompt by name (and optional version/label). - Action **Get Trace** Get a single trace by id, including nested observations/scores. - Action **List Dataset Items** List items in a dataset. - Action **List Datasets** List all datasets in the project. - Action **List Prompts** List all prompts in the project. - Action **List Scores** List scores with optional filters. - Action **List Traces** List traces with optional filters (user, session, tags, timerange). - Action **Create chat completion** Sends a conversation to an NVIDIA-hosted language model and returns the model's reply. - Action **Create embedding** Creates embedding vectors for one or more pieces of text using an NVIDIA NIM embedding model. - Action **Generate image** Generates an image from a text prompt with a FLUX model hosted on the NVIDIA API Catalog and returns it as base64. - Action **List models** Lists the models currently available for inference on the NVIDIA API catalog. - Action **Rerank documents** Rerank a list of candidate documents against a query for retrieval accuracy via an NVIDIA NIM reranking model. ## Each app on its own - [ **Langfuse** Langfuse is an open-source LLM engineering platform (GitHub) that helps teams collaboratively debug...](https://apiant.ai/connections/langfuse) - [ **Nvidia** API platform for deploying GPU-accelerated AI models, including inference microservices and a container...](https://apiant.ai/connections/nvidia) Reviews ## What our customers say ## More from Langfuse - [ **Langfuse to 5DollarFootball** Football fixtures, results, standings and historical odds with corner and card lines](https://apiant.ai/connect/langfuse-to-5dollarfootball) - [ **Langfuse to Cloudflare Trace** Cloudflare Trace is a free JSON wrapper over Cloudflare's edge trace endpoint that reports what Cloudflare...](https://apiant.ai/connect/langfuse-to-cloudflare-trace) - [ **Langfuse to PaperClean** Clean document photos for printing, remove shadows, fix lighting](https://apiant.ai/connect/langfuse-to-paperclean) - [ **Langfuse to Postman** Postman is a platform for building and testing APIs.](https://apiant.ai/connect/langfuse-to-postman) - [ **Langfuse to Rendobar** Hosted media processing over REST: transcode, caption, watermark video](https://apiant.ai/connect/langfuse-to-rendobar) - [ **Langfuse to Runpod** Cloud GPU platform for renting on-demand compute to train and deploy AI models, including serverless...](https://apiant.ai/connect/langfuse-to-runpod) ## More into Nvidia - [ **Google Drive to Nvidia** Google Drive is a cloud storage solution for uploading, sharing, and collaborating on files across devices...](https://apiant.ai/connect/google-drive-to-nvidia) - [ **Copper to Nvidia** Copper is a Recommended for Google Workspace CRM that improves productivity and increases visibility across...](https://apiant.ai/connect/copper-to-nvidia) - [ **MySQL to Nvidia** Connect directly to a MySQL database and run SQL statements from your automations.](https://apiant.ai/connect/mysql-to-nvidia) - [ **Clockify to Nvidia** Clockify is a completely free time tracking software for teams.](https://apiant.ai/connect/clockify-to-nvidia) - [ **HubSpot to Nvidia** HubSpot is your all-in-one stop for all of your marketing software needs.](https://apiant.ai/connect/hubspot-to-nvidia) - [ **Salesforce to Nvidia** Salesforce is a leading enterprise customer relationship management (CRM) platform.](https://apiant.ai/connect/salesforce-to-nvidia) Questions ## Langfuse and Nvidia, answered ### ### What can start an automation from Langfuse? Any of Langfuse's 1 trigger in the catalog, including New Trace. Describe the moment in plain language and your agent picks the right one. ### ### What can your agent do in Nvidia when something happens in Langfuse? Any of Nvidia's 5 actions, including Create chat completion, Create embedding, Generate image and List models. Your agent maps the data both ways and tests every scenario before it ships. ### ### Can it also run from Nvidia to Langfuse? Nvidia has no triggers in the catalog today, so automations start in Langfuse. If you need the other direction, ask: your agent reads Nvidia's API documentation and adds it. ### ### Do I have to build a connector for Langfuse or Nvidia? No. Both are in the catalog with their triggers and actions. You describe the automation; your agent builds it, tests every scenario and keeps it running. ### ### Can my AI client use Langfuse and Nvidia without building an automation? Yes. With your AI client connected to APIANT, ask it to do something in either app once and it runs that action on the spot. Ask for an automation and your agent builds it, tests every scenario and keeps it running. One connection covers both. ### ### What happens when something fails? Your agent works that out for you. It plans the failure paths along with the happy path, then tests every scenario before it ships. ### ### Is there a free plan? Yes. Every plan includes the agent, including free, and building from your own AI client spends no assistant credits.