
Humanloop
Your team runs dozens of prompt experiments and evaluation sessions in Humanloop. Now your AI agent can create projects, review experiment results, and surface session data on demand. LLM operations become conversational, and your product team stays focused on shipping.




Your AI agent interfaces directly with Humanloop's project and experiment APIs, turning complex LLM management tasks into simple conversational commands.
Humanloop
See how AI agents streamline Humanloop project management, experiment tracking, and session analysis for product teams building with language models.
An engineer messages 'Are all experiments passing for the onboarding project?' Your AI Agent calls Humanloop's List Experiments endpoint, scans each experiment's status and metrics, and responds with a summary showing three passed, one still running, and none failed. The engineer knows exactly where things stand before triggering a deploy.
A prompt engineer notices a regression in output quality. They ask the agent for recent sessions. Your AI Agent retrieves the latest session data from Humanloop, including inputs, outputs, and timestamps, and surfaces the last 10 interactions. The engineer spots the problematic pattern in minutes instead of digging through dashboards.
A product manager kicks off a new AI feature and needs a Humanloop project ready. They tell the agent 'Create a project called Customer Intent Classifier in the Support directory.' The agent creates the project with the right name, description, and directory assignment. The PM shares the project link with the team and evaluation work begins immediately.

Humanloop
FAQs
The agent uses Humanloop's REST API with your API key. It calls endpoints like Create Project, List Experiments, and List Sessions to retrieve or modify data. All requests are authenticated and scoped to your Humanloop organization. No data is stored by Tars between conversations.
You control this through agent configuration. Set up confirmation steps that require explicit approval before any delete operation. You can also restrict the agent to read-only operations like listing experiments and sessions, removing write access entirely if preferred.
Tars needs a Humanloop API key with access to the Projects and Experiments endpoints. For read-only setups, only list and retrieve permissions are needed. For full functionality including project creation and deletion, the key needs write access. You generate this key in your Humanloop account settings.
No. Tars queries Humanloop in real time during each interaction. Experiment statuses, session logs, and project details are fetched live. Nothing is cached or stored separately. Your evaluation data stays entirely within your Humanloop account.
Yes. The agent can query different project IDs within the same conversation. Ask about experiments in one project, then switch to session data from another. Each API call is scoped to the project ID you specify, keeping results clean and organized.
The dashboard requires you to navigate between projects, experiments, and sessions manually. A Tars AI Agent lets your team get answers through natural language, whether on Slack, WhatsApp, or your internal tools. Quick status checks take seconds instead of multiple clicks.
The agent receives an error response from Humanloop and handles it conversationally. It informs you that the project ID was not found, suggests verifying the ID format (must start with 'pr_'), and offers to list your available projects instead.
The agent can retrieve current experiment statuses and metrics on demand. For ongoing monitoring, configure it to check experiment results at regular intervals and alert your team when an experiment completes or fails. This turns Humanloop into a proactive notification system.
Don't limit your AI Agent to basic conversations. Watch how to configure and add powerful tools making your agent smarter and more functional.

Privacy & Security
At Tars, we take privacy and security very seriously. We are compliant with GDPR, ISO, SOC 2, and HIPAA.