different-ai/openwork · error

MCP_RESPONSE_BODY_LIMIT

MCP_RESPONSE_BODY_LIMIT

Error message

MCP_RESPONSE_BODY_LIMIT

What it means

MCP_RESPONSE_BODY_LIMIT is raised when the provider response exceeds the gateway's maximum body size, thrown as ExternalMcpResponseBodyLimitError and classified as response_too_large. The limit protects the gateway from unbounded payloads (huge tool catalogs, giant tool outputs, endless event streams). It is non-retryable and owned by the provider admin because the provider must shrink the payload.

Source

Thrown at ee/apps/den-api/src/capability-sources/external-mcp-diagnostics.ts:1191

  // The enterprise client aborts with a RequestTimeout MCP error of its own, so
  // check for our marker before any JSON-RPC code is read as the provider's.
  if (error instanceof ExternalMcpLifecycleDeadlineError || isEnterpriseMcpLifecycleDeadline(error)) {
    return {
      phase: fallbackPhase,
      category: "lifecycle_deadline",
      code: "MCP_LIFECYCLE_DEADLINE",
      retryable: true,
      actionOwner: "provider_admin",
      operatorAction: fallbackPhase === "MCP_TOOL_EXECUTION"
        ? "Retry the capability, and reduce provider latency for this tool if it keeps running past the bounded deadline."
        : "Reduce provider latency or catalog pagination so the complete MCP lifecycle finishes within the bounded deadline, then retry.",
    }
  }
  if (error instanceof ExternalMcpResponseBodyLimitError) {
    return {
      phase: fallbackPhase,
      category: "response_too_large",
      code: "MCP_RESPONSE_BODY_LIMIT",
      retryable: false,
      actionOwner: "provider_admin",
      operatorAction: "Reduce the provider response size, tool catalog, or event-stream payload before retrying.",
    }
  }
  if (error instanceof PrivateUrlError) {
    return {
      phase: "CONFIGURATION",
      category: "security_blocked",
      code: "MCP_URL_BLOCKED",
      retryable: false,
      actionOwner: "organization_admin",
      operatorAction: "Use a public HTTPS MCP URL or change the deployment's private-network policy through security review.",
    }
  }
  if (hasForbiddenPortMessage(error)) {
    return {
      phase: "CONFIGURATION",

View on GitHub (pinned to 2b7df46e8a)

Solutions

1. Reduce the provider response size — trim tool outputs to summaries or bounded result windows.
2. Shrink the exposed tool catalog (remove/deduplicate tools, paginate listing).
3. Cap event-stream payload sizes, then retry once the response fits within the body limit.

Example fix

// before: tool returns the entire dataset
return { rows: await db.select().from(events) }
// after: bounded output
return { rows: await db.select().from(events).limit(100), truncated: true }

// estimate payload size before sending large results through a tool
const serialized = JSON.stringify(result)
if (serialized.length > 1_000_000) {
  result = summarize(result) // truncate/summarize before returning from the tool
}

function isResponseBodyLimit(d: { code: string }): boolean {
  return d.code === 'MCP_RESPONSE_BODY_LIMIT'
}

try {
  return await client.callTool(req)
} catch (e) {
  if (isResponseBodyLimit(e.diagnostic)) {
    // non-retryable: request a narrower result window instead of retrying the same call
    return await client.callTool({ ...req, arguments: { ...req.arguments, limit: 100 } })
  }
  throw e
}

When it happens

Trigger: A tools/list response with an enormous catalog, a tool execution returning a very large result blob, or a text/event-stream that grows past the configured byte limit — any phase where the accumulated response body exceeds the limit before parsing completes.

Common situations: Providers exposing hundreds/thousands of tools in one unpaginated listing; tools returning entire datasets (full DB dumps, whole files) instead of summaries; misconfigured providers streaming verbose logs inside tool output; consumers requesting max-detail tool results.

Related errors

• MCP_PROVIDER_DECLARED_ERROR
• MCP_HTTP_404
• MCP_PROVIDER_HTTP_403
• MCP_HTTP_CONTENT_TYPE
• invalid_mcp_token_payload

AI-assisted analysis of different-ai/openwork@2b7df46e8a (2026-09-01). Data as JSON: /api/errors/a2de339a5ca03590. Report an issue: GitHub.