mastra-ai/mastra · error · TripWire

TokenLimiterProcessor: No messages to process. Cannot send L

Error message

TokenLimiterProcessor: No messages to process. Cannot send LLM a request with no messages.

What it means

TokenLimiterProcessor's `processInputStep` fetches all messages from the incoming MessageList and throws a TripWire (with `retry: false`) when there is nothing to process — an LLM request with zero messages is invalid. This guards against sending doomed requests downstream; the TripWire aborts the step rather than silently passing an empty list.

Source

Thrown at packages/core/src/processors/processors/token-limiter.ts:156

  /**
   * Process input messages at each step of the agentic loop, before they are sent to the LLM.
   * Runs at every step (including tool call continuations), preventing the conversation history
   * from growing unboundedly during multi-step agent workflows.
   *
   * System messages are always preserved, and the most recent non-system messages are kept
   * within the token budget.
   */
  async processInputStep(args: ProcessInputStepArgs): Promise<void> {
    const { messageList } = args;

    if (!messageList) return;

    const messages = messageList.get.all.db();

    // If no messages or empty array, throw TripWire - can't send LLM a request with no messages
    if (!messages || messages.length === 0) {
      throw new TripWire('TokenLimiterProcessor: No messages to process. Cannot send LLM a request with no messages.', {
        retry: false,
      });
    }

    // Budget against the full system message set that will reach the model
    // (untagged + tagged buckets), not just the untagged view exposed via args.
    const allSystemMessages = messageList.getAllSystemMessages();
    let systemTokens = 0;
    for (const msg of allSystemMessages) {
      systemTokens += await this.countCoreSystemMessageTokens(msg);
    }

    const limit = this.maxTokens;

    // If system messages alone exceed the token limit (accounting for conversation overhead),
    // throw TripWire - can't send LLM a request with only system messages
    if (systemTokens + TokenLimiterProcessor.TOKENS_PER_CONVERSATION >= limit) {
      throw new TripWire(

View on GitHub (pinned to 75dd419e61)

Solutions

1. Ensure at least one user (or system) message exists before invoking the pipeline.
2. Guard your caller: skip generation when `messages.length === 0` and return early instead.
3. Check whether an upstream processor (filter/redactor) is stripping all messages and adjust its rules.
4. Since `retry: false`, do not retry as-is — fix the input, or catch the TripWire and return a friendly 'no input' response.

Example fix

// before
await agent.generate(userInput?.trim() ?? '');
// after
if (!userInput?.trim()) return 'No input provided';
await agent.generate(userInput);

const msgs = messageList.get.all.db();
if (!msgs || msgs.length === 0) {
  // skip generation entirely instead of running the pipeline
  return null;
}

function hasMessages(list: MessageList | undefined): list is MessageList {
  return Boolean(list && list.get.all.db().length > 0);
}

try {
  result = await agent.generate(input);
} catch (e) {
  if (e instanceof TripWire && e.message.includes('No messages to process')) {
    result = { text: 'No input provided.' }; // retry: false, so don't retry
  } else throw e;
}

When it happens

Trigger: Calling an agent/generation step whose MessageList is empty: no user/system messages were added before the processor ran, all prior processors stripped every message, or a workflow step invokes the pipeline with an empty `messages: []` array.

Common situations: Building message lists dynamically where a filter removed all messages, empty user input passed straight through, memory/storage returning no messages for a fresh thread while the caller sends nothing new, and tests that construct agents without seed messages.

Related errors

• Aborted by processor
• Response body is null
• Response body is null
• execa is not available in Cloudflare Workers
• require(${specifier}) is not available in Cloudflare Workers

AI-assisted analysis of mastra-ai/mastra@75dd419e61 (2026-08-30). Data as JSON: /api/errors/7ad16d1d9af6a478. Report an issue: GitHub.