--- title: OpenPipe API Integration on Truto slug: openpipe category: Artificial Intelligence canonical: "https://truto.one/integrations/detail/openpipe/" --- # OpenPipe API Integration on Truto **Category:** Artificial Intelligence **Status:** Beta ## MCP-ready AI tools Truto exposes 23 tools for OpenPipe that AI agents can call directly. - **create_a_open_pipe_check_cach** — Check a cached prompt request in OpenPipe. Returns: respPayload. Deprecated: OpenPipe no longer supports prompt caching. Required: requestedAt. - **create_a_open_pipe_chat_completion** — Create a chat completion in OpenPipe. Returns the generated completion object including id, object, created, model, choices, and usage. Required: messages, model. - **create_a_open_pipe_report** — Record a request log from an OpenAI model in OpenPipe. Returns: status. - **create_a_open_pipe_report_anthropic** — Record a request log from an Anthropic model call in OpenPipe. Returns: status. Optional fields include statusCode, errorMessage, metadata, and tags. - **create_a_open_pipe_logs_update_tag** — Update tags on OpenPipe log entries matching the given filters. Returns: matchedLogs. Deprecated: use /logs/update-metadata instead. Required: filters, tags. - **create_a_open_pipe_logs_update_metadata** — Update tags metadata for logged OpenPipe calls matching the provided filters. Returns: matchedLogs. Required: filters, metadata. - **list_all_open_pipe_local_testing_only_get_latest_logged_calls** — Get the latest logged call from OpenPipe. Returns the call object with createdAt, cacheHit, statusCode, reqPayload, respPayload, tags, and metadata. Local testing only. - **create_a_open_pipe_criteria_judge** — Judge a completion against an OpenPipe criterion. Returns: score, explanation, usage. Required: criterion_id, output. - **list_all_open_pipe_datasets** — List datasets in an OpenPipe project. Returns dataset objects with id, name, created, updated, dataset_entry_count, and fine_tune_count. - **create_a_open_pipe_dataset** — Create a new dataset in OpenPipe. Returns the created dataset including id, name, created, updated, dataset_entry_count, and fine_tune_count. Required: name. - **delete_a_open_pipe_dataset_by_id** — Delete a dataset in OpenPipe by id. Returns the deletion status with id, object, and deleted. Required: id. - **create_a_open_pipe_dataset_entry** — Create new dataset entries in OpenPipe. Returns the creation summary including object, entries_created, and errors (with entry_index and message). Max 100 entries per request. Required: dataset_id, entries. - **list_all_open_pipe_models** — List all models in an OpenPipe project. Returns models with id, name, openpipe status, contextWindow, and pricing. - **create_a_open_pipe_model** — Train a new model in OpenPipe. Returns the created model including id, name, openpipe status, and pricing. Required: datasetId, slug, trainingConfig. - **delete_a_open_pipe_model_by_id** — Delete an OpenPipe model by its slug. Returns: id, object, deleted. Required: id. - **get_single_open_pipe_model_by_id** — Get a single OpenPipe model by id. Returns the model object including id, name, openpipe status, contextWindow, and pricing. Required: id. - **create_a_open_pipe_dataset_create** — Create a new dataset in openpipe. Returns: datasetId. Deprecated; prefer the `/datasets` endpoint. Required: name. - **create_a_open_pipe_dataset_delete** — Delete an OpenPipe dataset by its datasetId. Returns: datasetId. Deprecated endpoint; use the `/datasets/{dataset}` endpoint instead. Required: datasetId. - **list_all_open_pipe_dataset_lists** — List OpenPipe dataset lists via the deprecated /unstable/dataset/list endpoint. Returns an array of dataset objects with id, name, createdAt, updatedAt, datasetEntryCount, and fineTuneCount. Prefer the /datasets endpoint instead. - **create_a_open_pipe_dataset_entry_create** — Create dataset entries in OpenPipe. Returns: createdEntries, errors. Deprecated: use `/datasets/{dataset}/entries` instead. Required: datasetId, entries. - **create_a_open_pipe_finetune_create** — Create a fine-tune job in OpenPipe. Returns the new fine-tune job's id. Deprecated. Required: datasetId, slug, baseModel. - **list_all_open_pipe_finetune_gets** — Get a fine-tuned model in openpipe by id or slug. Returns: id, status, slug, baseModel, errorMessage, datasetId, createdAt. Deprecated: use the `/models/{model}` endpoint instead. - **create_a_open_pipe_finetune_delete** — Delete an openpipe fine-tuned model by id or slug. Returns: deleted. Deprecated: use the `/models/{model}` endpoint instead. ## How it works 1. **Link your customer's OpenPipe account.** Use Truto's frontend SDK; we handle every OAuth and API key flow so you don't need to create the OAuth app. 2. **Authentication is automatic.** Truto refreshes tokens, stores credentials securely, and injects them into every API request. 3. **Call Truto's API to reach OpenPipe.** The Proxy API is a 1-to-1 mapping of the OpenPipe API. 4. **Get a unified response format.** Every response uses a single shape, with cursor-based pagination and data in the `result` field. ## Use cases - **Offer end-users AI cost reduction on autopilot** — SaaS platforms that rely on GPT-4 or Claude for user-facing AI features can let their tenants automatically train cheaper, faster fine-tuned models from their own usage data. This turns AI infrastructure cost from a per-tenant liability into a competitive advantage. - **Enable tenant-specific model customization** — Multi-tenant SaaS products can give each customer their own fine-tuned model trained on that customer's edits, approvals, and domain data — without building an in-house ML pipeline. Every tenant gets a model that sounds like their business. - **Ship continuous learning from user feedback** — Product teams can wire Accept/Reject/Edit UI actions directly into OpenPipe's logging and metadata layer, so every user correction becomes labeled training data. This closes the loop between product usage and model quality automatically. - **Support regulated industries with private fine-tuned models** — LegalTech, healthcare, and fintech SaaS products can offer customers custom models trained on their sensitive data pipelines. This lets end-users meet compliance requirements without sacrificing the reasoning quality of frontier models. - **Build reliable multi-step agents with RL feedback** — SaaS platforms shipping AI agents can push success/failure signals back into OpenPipe as training data, improving agent reliability over time. This is critical for products where end-users judge quality by task completion, not just token output. ## What you can build - **Silent LLM telemetry capture** — Use the report endpoints to log every prompt-completion pair from your product's AI features into each tenant's OpenPipe workspace without altering user workflows. - **Feedback-driven metadata tagging** — Bind your in-app Accept, Reject, and Edit actions to the logs metadata update endpoint so user corrections instantly flag high-quality training examples in OpenPipe. - **Automated dataset curation pipelines** — Use dataset creation and dataset entry endpoints to programmatically batch cleaned, tagged conversation logs into per-tenant datasets when quality thresholds are hit. - **One-click fine-tuned model training** — Trigger training runs directly from your UI using the model creation endpoint, letting end-users kick off custom fine-tunes with their chosen base model and training config. - **Drop-in inference routing** — Once a tenant's fine-tune completes, route their inference calls through the OpenPipe chat completion endpoint to instantly cut token costs and latency for that customer. - **LLM-as-a-judge evaluation gates** — Call the criteria judge endpoint to score completions against custom rubrics before they enter a training dataset, ensuring only high-quality data reaches fine-tuning. ## FAQs ### How does authentication work for end-users connecting their OpenPipe account? End-users authenticate by providing their OpenPipe API key, which Truto securely stores and injects into every downstream request. Your app never needs to handle or persist the credential directly. ### Can we log completions from both OpenAI and Anthropic through this integration? Yes. Separate endpoints exist for reporting OpenAI-format logs and Anthropic-format logs, so you can pipe telemetry from either provider into the same tenant workspace. ### How do we attach user feedback signals to logged completions? Use the logs metadata update and tag update endpoints to attach arbitrary key-value metadata (e.g., user_accepted: true) or tags to any previously logged call, which can then be used to filter training data. ### Can we push pre-formatted training data directly into a dataset? Yes. The dataset entry creation endpoints let you push multi-turn chat examples into a specific dataset in batches, bypassing the log-and-curate flow if you already have clean data. ### How do we serve inference from a fine-tuned model once training completes? Call the chat completion endpoint with the fine-tuned model's slug. It follows the OpenAI-compatible request shape, so most existing chat completion code works with minimal changes. ### What lifecycle operations are supported for models and datasets? You can list, create, retrieve, and delete both models and datasets via dedicated endpoints, plus list finetune jobs and delete them — giving you full programmatic control over each tenant's training lifecycle.