Claude Honeycomb connectorWhat Claude can do in your Honeycomb account.
The Claude Honeycomb connector exposes 13 tools: 6 read, 1 write, 6 left undocumented. Claude walks through the environments, datasets, traces, SLOs and Triggers of your Honeycomb account for you, and can send feedback to the team that maintains the connector.
Verified Trustpilot reviews · AI, automation & growth agency
What changes when Claude reads your Honeycomb account
Mid-incident, you have no time to click from screen to screen. You describe the symptom to Claude in plain English, and it reads how your data is laid out, opens the trace in question and checks your reliability targets. It works with the access your consent opened to it when you connected, and nothing more.
Follow one slow request end to end. An on-call engineer pastes a trace ID. get_trace pulls every span as a waterfall, and Claude zooms into the subtree where the error shows up.
Check where your reliability targets stand. An SRE lead asks for the team's SLOs: get_slos lists them, or returns detailed status and graphs for one. get_triggers then shows who receives each alert.
Find the right field before digging in. get_dataset_columns returns a dataset's full schema, with sample values for the columns you ask about.
What it will not do: on this sheet, the only documented write tool sends feedback to Honeycomb. The publisher's docs describe other writes (Boards, Triggers, SLOs) that this sheet does not list, and six tools on the sheet are described nowhere. Nothing runs on its own either: Claude only acts while it answers you, and no alert wakes it up. For more examples, browse the Integrations hub.
Five terms before you connect
The vocabulary of this connector in one minute.
- Connector
- The link you set up once between Claude and an account you already own, so Claude can work on it while it answers you.
- Tool
- One named action a connector hands to Claude. Claude picks which ones to call by itself, and the directory lists them by name.
- Authorization
- Honeycomb's consent screen, where you decide what Claude may do; you grant it once and can take it back.
- Approval
- The confirmation Claude waits for before changing an account, shown in the conversation at the moment it matters.
- MCP
- The shared standard connectors run on: it is what lets an assistant like Claude talk to an outside service such as Honeycomb.
Connect Honeycomb to Claude in three steps
- 01
Find Honeycomb
Open Claude's settings, go to Customize, then Connectors, and search for Honeycomb. On a Team or Enterprise workspace, an Owner or Primary Owner has to switch it on for the organization first.
- 02
Sign in to Honeycomb
Click Connect: Honeycomb's sign-in window opens and you log in with your usual account. If the link ever breaks, Disconnect and a fresh connection put it right.
- 03
Read the consent screen
Honeycomb then shows what the access will cover. That screen, not Claude, sets the scope, so read it before you accept, especially the option that grants write access.
The 13 tools of the Honeycomb connector
Honeycomb gives Claude 13 tools: 6 that read your account, 1 that change something in it, and 6 no official source describes.
Honeycomb's official reference describes seven of these tools; six are not described anywhere. Names stay exactly as Claude shows them.
- 6 read
- 1 write
- 6 not documented
What Claude reads (6)
6 toolsSix reads from your Honeycomb account, changing nothing.
get_dataset
Opens one dataset and gives its metadata plus the full schema: columns and calculated fields in a single list, sorted by when they were last written to.
get_dataset_columns
Returns a dataset's full column schema and, on request, sample values for specific columns. For a metrics dataset, it returns metric names by default.
get_environment
Describes one environment, with its datasets and calculated fields ranked by recent activity. Honeycomb returns up to 100 datasets per call.
get_slos
Lists the team's SLOs, or shows detailed status and graphs for one particular SLO, so you can see at a glance where each reliability objective stands.
get_trace
Pulls every span for a trace ID and lays them out as a waterfall. Claude can zoom into a subtree to isolate where the error happens.
get_triggers
Goes through the team's Triggers, or shows the full configuration of one, recipients included: type, name or target, and ID.
What Claude can send (1)
1 toolA single write tool, which needs the write right granted at consent.
feedback
Approval: see the rulePasses your opinion of the connector to the Honeycomb team that maintains it. You simply ask Claude to send feedback to Honeycomb along with your message.
Undocumented (6)
6 toolsThe directory publishes these six names and nothing more. Honeycomb's reference does not list them, and we will not guess their role from a name.
get_query_history
This name sits in the directory's tool list and that is all we know. The tools reference Honeycomb publishes, which walks through each of its tools, never mentions it, and no Claude page covers it either.
get_team_info
Same silence here: the name is on the sheet, but no page from the publisher or from Claude describes it. Rather than guess from how it sounds, this page flags the gap and stops there.
list_datasets
This name appears on the directory sheet but not in the official list Honeycomb publishes, even though that list covers its tools one by one with their role. No other source fills the gap.
list_environments
The directory gives its name and nothing else. Neither Honeycomb nor Anthropic had published a description at the time of our review, and this page does not go beyond that, however telling the name may look.
run_unsaved_query
No official source says what this tool does, or whether it only reads or also changes something in your account. Only its name is published on the directory sheet, with no sentence to go with it.
search_columns
The last name on the sheet that nobody documents. Honeycomb's reference does not list it, and the directory only publishes its label.
What Claude asks before it acts
By default, Claude stops and asks for your go-ahead before each action it takes on an account for someone. With Honeycomb, writing also depends on a right you grant during consent.
No source describes a confirmation specific to feedback, so the default rule is the reference, as it is for the six undocumented tools. On a Team or Enterprise workspace, owners decide whether a member may let some actions through without confirming, and can cap a connector for the whole organization. On the publisher's side, the Activity Log environment is surfaced to assistants only for the owners of the Honeycomb team. Claude never gets more than the person who connected it.
Which Claude plans include Honeycomb
None of the 819 sheets in the official directory shows availability by plan. Connector by connector, that information is not published anywhere.
The general rule is sourced: remote connectors are open to all users on Claude, Cowork, Claude Desktop and mobile. On Team and Enterprise, an Owner or Primary Owner enables the connector before members can sign in to it. For Honeycomb's current state, check its sheet in the official directory, the only place that shows it for your own account.
Where the Honeycomb connector stops
A connector is not an automation. Claude calls these tools while it answers you, so a firing Trigger does not bring it in.
The sheet is only a floor: Honeycomb's reference describes tools, including writes on Boards, Triggers and SLOs, that do not appear in this list, and six names on the sheet are missing from it. Only the publisher, Honeycomb, documents this connector; Anthropic has no page of its own on it. The directory badge is not a security audit, and Anthropic does not guarantee tools behave exactly as described.
Need help connecting Honeycomb to Claude?
A person reads every message.