canonical: https://jentic.com/apis/helicone.ai/helicone

# Helicone API

The Helicone API logs and analyzes large language model requests for LLM applications. It queries logged requests and their inputs, attaches scores, feedback, and custom properties, groups calls into agent sessions with metrics, manages prompts and prompt versions, runs prompt experiments and evaluators, and administers provider keys and API keys.

## For AI agents

Query logged LLM requests, score and annotate them, group calls into sessions, manage prompts and experiments, and run evaluators. Authenticates with an API key sent as a Bearer token in the Authorization header.

## Scope

Does not run model inference itself. Use the Helicone API to log, query, and analyze LLM usage and to manage prompts, experiments, and keys.

## Capabilities

- Query logged LLM requests and their inputs
- Attach scores, feedback, and custom properties to requests
- Query agent sessions and their metrics
- Manage prompts and prompt versions
- Run and query prompt experiments and evaluators
- Manage provider keys and API keys

## Use cases

### LLM Request Monitoring

An AI agent connected through Jentic can inspect how an LLM application is behaving. It queries logged requests, reads their inputs and scores, and pulls request metrics, so an assistant can surface error rates, latency, or cost trends without a separate dashboard session.

Example prompt: Query the most recent logged requests and summarize their error rate

### Prompt Experimentation

Teams iterating on prompts need to compare versions against data. The Helicone API creates experiments, manages datasets, and runs evaluators over them, letting an agent set up a comparison and read back the scored results.

Example prompt: Create an experiment and run an evaluator over its dataset

### Session Analytics

Multi-step agents produce sessions rather than single calls. The Helicone API queries sessions, their names, and aggregate metrics, so an agent can report on how a conversation or workflow performed end to end.

Example prompt: Query sessions and return their aggregate metrics

## Key endpoints

| Method | Path | Description |
| --- | --- | --- |
| POST | `/v1/request/query` | Query logged LLM requests |
| GET | `/v1/request/{requestId}` | Get a logged request by ID |
| POST | `/v1/session/query` | Query agent sessions |
| POST | `/v1/experiment/query` | Query prompt experiments |
| POST | `/v1/prompt-2025/query` | Query prompts |
| GET | `/v1/models` | List models |

## Key resources

- **Requests** — Logged LLM requests, their inputs, scores, and feedback
- **Sessions** — Grouped agent sessions and their metrics
- **Prompts** — Prompts and their versions
- **Experiments** — Prompt experiments, datasets, and tables
- **Evaluators** — Evaluators run against requests and experiments
- **Keys** — Provider keys and API keys

## Why Jentic

- **Setup:** Wiring Helicone by hand means creating an API key, building request and session queries, and paging through large result sets yourself. Through Jentic you install once, import Helicone from the API Directory, store the key once, and your agent calls the query and management operations directly.
- **Permission scoping:** You choose which of the operations your agent may call and can keep it to read-only queries. You might allow request and session queries while blocking key management and prompt edits.
- **Credential handling:** Your Helicone API key is stored once, encrypted, by your own Jentic One instance and injected at execution time. It never enters the agent's prompt, logs, or context.
- **Discovery method:** Agents search Jentic by intent such as 'query logged LLM requests', and Jentic returns the matching Helicone operation with its input schema so the agent calls it with the right filters.

## Related APIs

- **Langfuse** — Open-source LLM observability with traces, prompts, and evaluations
- **OpenAI API** — Chat completions, embeddings, and other model endpoints
- **New Relic** — Application performance monitoring and observability platform

## FAQ

### What authentication does the Helicone API use?

Most operations authenticate with an API key sent as a Bearer token in the Authorization header, per the OpenAPI spec. A few public endpoints, such as the model list, need no key. Through Jentic the key is stored once, encrypted, by your own self-hosted instance and injected when the agent calls, so it never appears in the agent's prompt or logs.

### Can I limit what my agent is allowed to do with the Helicone API?

Yes. You choose which operations the agent may call and can keep it to read-only queries. You might allow request and session queries while blocking key management and prompt edits, with every call logged by your own instance.

### Is there a Helicone MCP server?

You don't need an MCP server to give your agent the Helicone API. Jentic connects it directly from the API Directory: import it, store your key once, and your agent calls the request, session, and experiment operations. Operations are discovered on demand, so nothing extra loads into the agent's context.

### What are the rate limits for the Helicone API?

The OpenAPI spec does not define rate limits. Check the Helicone documentation for current plan limits before running high-volume queries against the request logs.

### What can I do with the Helicone API?

You can query logged LLM requests and their inputs, attach scores and feedback, group calls into sessions with metrics, manage prompts and their versions, run experiments and evaluators, and administer provider and API keys.

### How do I query LLM request logs through Jentic?

Search Jentic by intent, for example 'query logged LLM requests', and Jentic returns the matching Helicone operation with its input schema. Your agent supplies the filters and reads back the matching requests. To run it on your own infrastructure, install Jentic One from its GitHub repo.