---
title: "Helicone MCP Server: review, setup & limitations · Undominated.ai"
canonical: https://undominated.ai/mcp-servers/helicone/
description: "Connects an assistant to Helicone request logs and sessions, and can send new LLM requests through the Helicone AI Gateway. It runs locally over stdio via the published npm package with a Helicone API key."
---

# Helicone MCP Server: review, setup & limitations · Undominated.ai

> Connects an assistant to Helicone request logs and sessions, and can send new LLM requests through the Helicone AI Gateway. It runs locally over stdio via the published npm package with a Helicone API key.

[← Explore all mcp servers](/mcp-servers/)

LLM OBSERVABILITY / Helicone

# Helicone MCP Server

Connects an assistant to Helicone request logs and sessions, and can send new LLM requests through the Helicone AI Gateway. It runs locally over stdio via the published npm package with a Helicone API key.

 [See setup guidance ↓](#setup)[Original source ↗](https://github.com/Helicone/helicone)

SOURCE REVIEW

 Reviewed 2026-10-13
 Evidence 3 linked sources
 Publisher Helicone
 Licence [Apache-2.0 ↗](https://github.com/Helicone/helicone/blob/067d9290acb4f1fc9320e902fc67b4b399b50363/LICENSE)
 Revision 067d9290acb4
 [Read what was—and wasn’t—checked ↓](#review)

“A Model Context Protocol (MCP) server for querying Helicone observability platform data and making LLM requests through the Helicone AI Gateway.”

 [Helicone · upstream description ↗](https://github.com/Helicone/helicone/blob/067d9290acb4f1fc9320e902fc67b4b399b50363/helicone-mcp/README.md) Our analysis follows below.

01 / THE REASONING

## Why this made the selection.

 - Publisher-maintained observability surface: query_requests and query_sessions filter logged traffic by model, status, cost, latency and time without leaving the assistant.
- Ships as a versioned npm package with a documented stdio configuration at the pinned revision.

### A good fit for

 - Tracing a failed or expensive model call back through its request log.
- Summarising spend and error patterns across a Helicone project.

### Weigh up before choosing

 - Reads Helicone-logged traffic only; requests that bypass the Helicone gateway are invisible to it.
- Needs a Helicone API key. The same key that reads logged request data also sends new requests through use_ai_gateway.
- use_ai_gateway sends new LLM requests to model providers through the Helicone AI Gateway, so this server is not read-only.
- query_requests can return request and response bodies (includeBodies), which places logged prompt and completion text in the assistant context.

02 / THE REVIEW RECORD

## What we actually inspected.

Source review has boundaries. A clear record is more useful than a “safe” badge.

### Material inspected

 - helicone-mcp/README.md at 067d9290acb4f1fc9320e902fc67b4b399b50363: tools (use_ai_gateway, query_requests, query_sessions), stdio configuration and required API key.
- Package manifest: runtime and published entrypoint at the pinned revision.
- LICENSE: Apache-2.0 terms for the repository source.

### Our findings

 - The documented configuration is stdio-only via npx with the API key passed as an environment variable.
- Corrected 2026-10-13: the README lists three tools, use_ai_gateway, query_requests and query_sessions. use_ai_gateway makes LLM requests through the Helicone AI Gateway, so the earlier statement that the server sends no model requests was wrong.
- The configExample shows YOUR_HELICONE_API_KEY where the README block prints an example key value. It is otherwise the README block as printed.
- helicone-mcp/package.json at this commit names @helicone/mcp, matching the install command. HEAD on 2026-10-13 equals the pinned commit.

### Not established by this review

 - Server startup, authentication and tool execution were not tested.
- use_ai_gateway was not called; no model request was sent.

The review applies to the material and revision named here. A newer upstream release can change its behavior.

03 / PUT IT TO WORK

## Connect a server deliberately.

[Upstream setup instructions ↗](https://github.com/Helicone/helicone/blob/067d9290acb4f1fc9320e902fc67b4b399b50363/helicone-mcp/README.md)

DOCUMENTED COMMAND

 npx @helicone/mcp@latest Copy command ↗

Copying does not execute this command. It may retrieve a newer version than the reviewed source.

 - Add the documented stdio configuration to the MCP client with the Helicone API key in the environment.
- Confirm the key reaches only the intended Helicone project before querying logs.
- The command above is the command and args of the README’s stdio block (npx with @helicone/mcp@latest), not a separate installer. It was not run.

### Before you start

 - Node.js with npx available.
- A Helicone API key with access to the target project.

### Documented configuration example

Use the client and authentication context described in the setup guidance. Configuration formats vary between hosts.

 {
 "mcpServers": {
 "helicone": {
 "type": "stdio",
 "command": "npx",
 "args": [
 "@helicone/mcp@latest"
 ],
 "env": {
 "HELICONE_API_KEY": "YOUR_HELICONE_API_KEY"
 }
 }
 }
} Copy configuration ↗

### Compatibility

MCP clients supporting stdio with Node.js and npx available. · Helicone account with logged request traffic in the target project.

### Implementation

TypeScript

### Transports

stdio

### Helicone API key

 - Reads request logs and sessions visible to the configured key, optionally including request and response bodies.
- use_ai_gateway issues new LLM requests through the Helicone AI Gateway under the configured key. The README lists no tool that changes project settings.

### Cost model

Helicone account entitlements apply; the local source licence does not include a Helicone plan.

04 / FOLLOW THE EVIDENCE

## The source trail.

Our notes are separate from the original resource. Check upstream before adopting a new version.

 - [Helicone MCP server readme ↗](https://github.com/Helicone/helicone/blob/067d9290acb4f1fc9320e902fc67b4b399b50363/helicone-mcp/README.md) Checked 2026-10-13 https://github.com/Helicone/helicone/blob/067d9290acb4f1fc9320e902fc67b4b399b50363/helicone-mcp/README.md Supports: summary, install, compatibility, access, transports, configExample, whySelected, bestFor, limitations, upstreamDescription, Tool list including use_ai_gateway, includeBodies option of query_requests, Filter fields of query_requests and query_sessions
- [Repository licence ↗](https://github.com/Helicone/helicone/blob/067d9290acb4f1fc9320e902fc67b4b399b50363/LICENSE) Checked 2026-10-13 https://github.com/Helicone/helicone/blob/067d9290acb4f1fc9320e902fc67b4b399b50363/LICENSE Supports: license
- [Helicone MCP package manifest ↗](https://github.com/Helicone/helicone/blob/067d9290acb4f1fc9320e902fc67b4b399b50363/helicone-mcp/package.json) Checked 2026-10-13 https://github.com/Helicone/helicone/blob/067d9290acb4f1fc9320e902fc67b4b399b50363/helicone-mcp/package.json Supports: Package name @helicone/mcp, Published entrypoint build/index.js

KEEP COMPARING

## Other approaches to consider.

Related by category or shared topics. These are alternatives to inspect, not a measured quality order.

 [### MLflow MCP Server ↗ Ships inside the mlflow package and exposes trace search, feedback and assessment logging, and optionally experiment, run, model and deployment commands, to an MCP client.](/mcp-servers/mlflow-mcp/)[### OpenTelemetry MCP Server ↗ Queries LLM application traces in Jaeger, Grafana Tempo or Traceloop, with tools that aggregate token usage by model and find slow, token-heavy or failing traces.](/mcp-servers/opentelemetry-mcp/)[### Opik MCP Server ↗ Reads and scores Opik traces, experiments, datasets and prompts from an MCP client. Comet hosts it for Opik Cloud; self-hosted and open-source Opik run the same package locally over stdio.](/mcp-servers/opik-mcp/)

 [AI Tools ↗](/tools/)[Skills ↗](/skills/)[Agents ↗](/agents/)[MCP Servers ↗](/mcp-servers/)[Workflows ↗](/workflows/)

## Continue your investigation

 - [Inspect resource evidence](/tools/)
- [Choose a host](/agents/)
- [Inspect reusable workflows](/skills/)
- [Use Undominated data](/api/)
