Honeycomb MCP server icon

Honeycomb

by Honeycomb

HIPAA CompliantSOC2 ReadyISO 27001 Ready
Data & Research13 tools

Query observability data, pull distributed traces and inspect SLOs from your AI agent. Honeycomb documents 34 tools including eight writes behind an mcp:write scope, OAuth sign-in, and a team-level AI switch that is off until an owner turns it on.

Verified connector

Listed by Anthropic as a partner connector in its Connectors Directory.

Connection checked by Agentman on .

Anthropic states this reflects the level of review a connector received, not a security audit.

Connect Honeycomb via MCP

https://mcp.honeycomb.io/mcp

Works in any MCP-compatible client. In Agentman Studio it is one click — no config file to edit.

Use in Agentman

Connect Honeycomb once and your agents call these tools on their own — on a schedule, in a workflow, with nobody at the keyboard.

Open in Agentman Studio

Honeycomb Tools & Capabilities (13)

feedback
get_dataset
get_dataset_columns
get_environment
get_query_history
get_slos
get_team_info
get_trace
get_triggers
list_datasets
list_environments
run_unsaved_query
search_columns

Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.

Limits

  • The directory tool list is stale. Thirteen names against 34 documented, five of them renamed. Any integration built from the snapshot will call tools that no longer exist.
  • We could not verify tool behaviour ourselves. The endpoint returns 401 to an anonymous handshake, so no readOnlyHint or destructiveHint annotation was readable. Read/write classification here is Honeycomb's, from its documentation, not ours from the wire.
  • Team Owners hold two switches. Honeycomb Intelligence gates the whole connector, and only Team Owners can create the Management API keys used for unattended agents.
  • Rate limits are per tool and can bite mid-investigation. get_service_map allows 10 calls a minute, the create tools 20, update_trigger and update_slo 30. Honeycomb notes limits are subject to change and suggests upgrading a plan if you hit them regularly.
  • No tool reports remaining allowance. Nothing in the documented surface returns a rate-limit budget or an ingest figure, so an agent discovers a limit by hitting it.
  • Sessions expire after about 24 hours, and Honeycomb's own troubleshooting page concedes some OAuth clients behave inconsistently across platforms, directing affected users to support.
  • The Activity Log Environment is withheld from agents for everyone except Team Owners, even though other roles can see it in the Honeycomb UI.
  • Two tools carry extra gates: get_service_map is Enterprise-only, and get_anomaly_service_profiles requires anomaly detection enabled for the team.
  • No redaction layer. Honeycomb documents no MCP-side masking of span attributes, and its AI terms place the duty to keep regulated data out of Honeycomb Intelligence on the customer.

Frequently asked questions

It can change things. Honeycomb's own tools reference documents eight write tools that create or update Boards, Triggers, SLOs, notification recipients and Canvas investigations, plus a `feedback` tool that sends a message to Honeycomb. Every one of them requires the `mcp:write` scope, which you grant separately during OAuth consent or on an API key. Read tools need only `mcp:read`.

The directory snapshot is stale, and Honeycomb's documentation is the current surface. Anthropic's 2026-08-16 snapshot names 13 tools, five of which no longer appear in Honeycomb's tools reference under those names. Honeycomb's page documents 34 tools across twelve categories, 26 of them read-only. Treat the directory list as a historic subset and the vendor page as authoritative.

Honeycomb Intelligence is probably disabled for your team. Honeycomb's troubleshooting page says authenticating successfully but seeing no tools, or having every call fail on authorization, usually means the team-level AI setting is off. Only a Team Owner can enable it in team settings. That switch gates the whole connector, not individual tools.

Yes, if that data is in your telemetry. `get_trace` returns every span for a trace ID as a waterfall, and `get_dataset_columns` can return sample values. Honeycomb's AI policies page states MCP uses sample telemetry values, and its AI terms forbid customers from putting protected health information or GDPR special-category data through Honeycomb Intelligence at all.

No. Honeycomb prices its plans by events ingested and metrics data points, not by queries run, so reading through MCP does not consume ingest allowance. Queries are instead governed by per-tool rate limits: 200 or 300 calls a minute for discovery tools, 100 for `run_query` and `get_trace`, and as low as 10 for the Enterprise-only service map.

The documentation no longer says so. Anthropic's directory points at a URL under a `product-lifecycle/beta/` path, but that URL now redirects to `/integrations/mcp/concepts`, a permanent integrations page. Nothing in Honeycomb's four current MCP pages carries a beta, preview or experimental label, and the pricing page lists Honeycomb MCP as a Free-plan feature.

The Free plan is enough. Honeycomb's pricing page lists Honeycomb MCP among Free-tier features, alongside distributed tracing, BubbleUp and the Canvas AI copilot, with an ingest ceiling of 20 million events a month. Two tools carry extra gates: `get_service_map` is badged Enterprise, and `get_anomaly_service_profiles` needs anomaly detection enabled for the team.

Sources

  • Honeycomb MCP Core Concepts — https://docs.honeycomb.io/integrations/mcp/concepts (fetched 2026-08-18 as Markdown via the .md suffix; the directory's documentation URL docs.honeycomb.io/troubleshoot/product-lifecycle/beta/mcp/ returns 308 to /integrations/mcp, which returns 307 to this page) · retrieved 2026-08-18
  • Honeycomb MCP Tools Reference — https://docs.honeycomb.io/integrations/mcp/tools (fetched 2026-08-18) · retrieved 2026-08-18
  • Connecting to Honeycomb MCP — https://docs.honeycomb.io/integrations/mcp/configuration-guide (fetched 2026-08-18) · retrieved 2026-08-18
  • Troubleshooting Honeycomb MCP — https://docs.honeycomb.io/integrations/mcp/troubleshooting (fetched 2026-08-18) · retrieved 2026-08-18
  • Honeycomb AI Policies — https://docs.honeycomb.io/security-compliance/ai-policies (last updated 16 March 2026; fetched 2026-08-18) · retrieved 2026-08-18
  • Honeycomb Intelligence Supplemental Terms — https://www.honeycomb.io/ai-terms (fetched 2026-08-18) · retrieved 2026-08-18
  • Honeycomb pricing — https://www.honeycomb.io/pricing (fetched 2026-08-18) · retrieved 2026-08-18
  • Live RFC 9728 protected-resource descriptor — https://mcp.honeycomb.io/.well-known/oauth-protected-resource (fetched 2026-08-18; the /mcp-suffixed form returns a byte-identical body) · retrieved 2026-08-18
  • Live RFC 8414 authorization-server metadata — https://ui.honeycomb.io/.well-known/oauth-authorization-server (fetched 2026-08-18) · retrieved 2026-08-18
  • Live anonymous probe — 2026-08-18: POST returns HTTP 401 with www-authenticate: Bearer resource_metadata=…; GET returns 405 · retrieved 2026-08-18
  • docs.honeycomb.io/robots.txt disallows only /cdn-cgi/ and /_next/, names a sitemap, and carries Content-Signal: ai-train=yes, search=yes, ai-input=yes — synthesis is expressly permitted on all three axes (fetched 2026-08-18) · retrieved 2026-08-18
  • Honeycomb docs machine index — https://docs.honeycomb.io/llms.txt (fetched 2026-08-18) · retrieved 2026-08-18
  • Anthropic connector directory — https://claude.ai/directory/63125588-06e1-4567-8130-f5a26ed1d20f (snapshot 2026-08-16) · retrieved 2026-08-16

Server Info

Category
Data & Research
Developer
Honeycomb
Tools
13
Domain
mcp.honeycomb.io

Using Claude Desktop or another MCP client? Setup docs — the connection URL above works anywhere.

Ready to connect Honeycomb?

Connect Honeycomb once and your agents call these tools on their own — on a schedule, in a workflow, with nobody at the keyboard.