Runheed

What Runheed does

Runheed brings production error data into MCP-compatible coding agents. It collects exceptions from Node.js and Python applications and organizes them into error groups that agents can query by text, status, and project. Each group includes identifying information such as its type, message, occurrence count, timestamps, and status.

The service is intended for debugging workflows where an agent needs more than a log line. It can expose stack frames, breadcrumbs, runtime and platform details, release and environment information, affected users, agent sessions, and additional recorded data. Agents can also inspect workspace or project health, including recent error activity, unresolved issues, error rate, trends, and recommendations.

How it works

Runheed exposes its functionality through a hosted MCP endpoint protected with OAuth. An MCP client connects to that endpoint, allowing an agent to call error-tracking tools during a coding session. The agent can search for an issue, retrieve its full details or surrounding context, examine recent activity and longer-term statistics, and update the issue's state.

Resolved errors can reopen automatically when they occur again. Ignored errors remain excluded from new alerting until they are moved back to an unresolved state. A resolution can include an optional note, such as a fixing commit, to preserve a link between the issue and the change that addressed it.

Tools and capabilities

• Search error groups by message or type, status, and project.
• Retrieve a complete error group with counts, timestamps, environment, agent context, stack frames, and recent breadcrumbs.
• List the most recently seen error groups.
• Review statistics for the previous 24 hours, 7 days, or 30 days.
• Retrieve all recorded context surrounding an error.
• Mark errors resolved, optionally adding a note.
• Mark known or expected errors ignored, or restore them to unresolved.
• Check a project or workspace health score from 0 to 100.

The trend data includes total, new, and resolved counts, frequent errors, and hourly error rate. Project health also includes recent issues and recommendations.

Setup and configuration

Runheed is accessed as a hosted MCP service rather than as a local process. Configure an MCP-compatible client to connect to its OAuth-protected endpoint, then authorize access. The service is designed for Claude Code, Cursor, Codex, and other MCP clients.

The monitored application scope is Node.js or Python. Once application exceptions are available to Runheed, the MCP tools expose them to the connected agent for investigation and status updates. Paid plans start at $29 per month.

Limitations and notes

Runheed focuses on captured application exceptions and their associated diagnostic context. Its documented application coverage is Node.js and Python. Actions that change issue state include resolving, ignoring, and unignoring error groups; these changes affect how subsequent occurrences are handled. Resolved groups reopen if the same error happens again, while ignored groups stop generating alerts until restored.