Integration

HoneyHive integration

HoneyHive is an observability and evaluation platform for LLM applications and agents. With a project API key you can log traces (start sessions, create events singly or in batches, update events with feedback, metrics and metadata), search and read logged events, and manage evaluators (called metrics in the API): create, update, version, deploy and test-run them on an event. Datasets, datapoints, charts, experiment runs and annotation queues need a different key and are in the HoneyHive Data Plane integration; project, workspace and alert management (control plane) and API key creation are not exposed, and the deprecated legacy routes are left out. Authentication: a classic project API key (starts with hh_, or hh_ro_ for read-only), created in HoneyHive under Settings > Project > API Keys on the Project tab, sent as Authorization: Bearer. The key fixes the project. Calls go to the default US data plane https://api.dp1.us.honeyhive.ai.

Connect HoneyHive How integrations work

Ask it in chat

Say what should happen.
The workflow builds itself.

When this happens, do that in HoneyHive.Built from the 15 HoneyHive tools below, on the connection you add once.

What you can do with HoneyHive

Each of these is one operation on one endpoint, available both as a workflow step and as a tool an AI agent can call on its own. Nothing bundles several operations into one action, so what a run did stays readable afterwards.

Most endpoints run on the free plan. A few, mainly the ones that publish on your behalf, need a paid plan.

Event, 6 operations

create_eventCreate eventPOSTLogs one event (span) in a session trace of the key's project and returns its event_id. event_type (chain, model, tool or session) and inputs are required; pass session_id to attach it to a session.
create_event_batchCreate event batchPOSTLogs several events at once and returns their event_ids. With single_session true, all of them go into one new session built from session_properties, whose session_id is returned.
create_session_event_batchCreate session event batchPOSTAdds several events (spans) to an existing session; each event gets the session_id of the path. Each event needs event_type (chain, model, tool or session) and inputs.
get_eventGet eventGETReturns one logged event (span or session) by id, with its inputs, outputs, config, metadata, feedback, metrics, timing and parent/children ids.
search_eventsSearch eventsPOSTSearches the project's logged events with filters and a date range, newest first (by start_time). Returns events and count. Page with limit (default 1000) and page (from 1).
update_eventUpdate eventPUTUpdates an existing event: metadata, feedback, metrics, config and user_properties are MERGED into what it has; outputs is replaced. Fields left out are unchanged. Use it to attach user feedback or scores later.

Metric, 5 operations

create_metricCreate metricPOSTCreates an evaluator (metric) in the project: a Python check, an LLM judge prompt or a human annotation field. Returns metric_id. With enabled_in_prod true it starts scoring production traces.
delete_metricDelete metricDELETEPermanently deletes an evaluator (metric) from the project.
list_metricsList metricsGETLists the project's evaluators (called metrics): name, type, criteria, return type, threshold, production sampling and filters. Filter by type or id.
run_metricRun metricPOSTRuns an evaluator definition you pass on one event you pass and returns result and explanation. An LLM evaluator calls its model with the workspace's provider credentials, which that provider bills.
update_metricUpdate metricPUTUpdates an evaluator (metric). Only the fields you send change; the others keep their current value.

Metric version, 3 operations

create_metric_versionCreate metric versionPOSTSaves a new version of an evaluator's definition. It is a draft unless deploy_immediately is true, which also makes it the live version.
deploy_metric_versionDeploy metric versionPOSTMakes the named version the live version of the evaluator, replacing the previously deployed one. Affects how production traces are scored.
list_metric_versionsList metric versionsGETLists every saved version of an evaluator's definition, oldest first, unpaged: version name, message, date, whether it is deployed, and the definition.

Session, 1 operation

create_sessionCreate sessionPOSTStarts a new trace session in the key's project and returns the created session event (event_id, session_id). Pass session_id to choose the id yourself; every field is optional.

Use HoneyHive from Claude Code, Cursor or Codex

Connect HoneyHive in LiveContext, create an API key in Settings, then add the LiveContext MCP server to your client. Its HoneyHive operations become tools the client can call, next to every other integration you connected.

Claude Code

claude mcp add --transport http livecontext https://livecontext.ai/mcp --header "X-API-Key: YOUR_API_KEY"

Codex (~/.codex/config.toml)

[mcp_servers.livecontext]
url = "https://livecontext.ai/mcp"

[mcp_servers.livecontext.http_headers]
"X-API-Key" = "YOUR_API_KEY"

Cursor, Claude Desktop and other JSON configs

{
  "mcpServers": {
    "livecontext": {
      "type": "http",
      "url": "https://livecontext.ai/mcp",
      "headers": {
        "X-API-Key": "YOUR_API_KEY"
      }
    }
  }
}

Set up the MCP server step by step

How to connect HoneyHive

Create your workspace
Sign up free. Nothing to install.
Connect HoneyHive once
Add your HoneyHive connection in Settings. Every tool from this integration reuses it, and secrets are encrypted at rest and visible only to you.
Describe the job in chat
The workflow builds itself in front of you and picks the endpoints it needs. Drag the steps yourself on the canvas whenever you would rather be explicit.